redgpu/raygpu: Simple & Portable Graphics Library

src/models.c
2 970 строк133 KB
manuel
Work on things
07 сен 2025, 19:22
07 сен 2025, 19:225cacb1d
#include <raygpu.h>
#ifndef RL_CALLOC
#define RL_CALLOC calloc
#endif
#ifndef RL_MALLOC
#define RL_MALLOC malloc
#endif
#ifndef RL_REALLOC
#define RL_REALLOC realloc
#endif
#ifndef RL_FREE
#define RL_FREE free
#endif
#ifndef RL_REALLOC
#define RL_REALLOC realloc
#endif
#define TINYOBJ_LOADER_C_IMPLEMENTATION
#include <external/tinyobj_loader_c.h>
#include <external/cgltf.h>
#define PAR_SHAPES_IMPLEMENTATION
#include <external/par_shapes.h>
#ifndef MAX_MATERIAL_MAPS
    #define MAX_MATERIAL_MAPS       12    // Maximum number of maps supported
#endif
#ifndef MAX_MESH_VERTEX_BUFFERS
    #define MAX_MESH_VERTEX_BUFFERS  9    // Maximum vertex buffers (VBO) per mesh
#endif


const float v3_zero [3] = {0,0,0};
const float v3_xunit[3] = {1,0,0};
const float v3_yunit[3] = {0,1,0};
const float v3_zunit[3] = {0,0,1};
const float v3_xunit_negative[3] = {-1,0,0};
const float v3_yunit_negative[3] = {0,-1,0};
const float v3_zunit_negative[3] = {0,0,-1};

cgltf_result LoadFileGLTFCallback(const struct cgltf_memory_options *memoryOptions, const struct cgltf_file_options *fileOptions, const char *path, cgltf_size *size, void **data){
    size_t filesize;
    void* filedata = LoadFileData(path, &filesize);

    if (filedata == NULL) return cgltf_result_io_error;

    *size = filesize;
    *data = filedata;

    return cgltf_result_success;
}

// Release file data callback for cgltf
static void ReleaseFileGLTFCallback(const struct cgltf_memory_options *memoryOptions, const struct cgltf_file_options *fileOptions, void *data){
    UnloadFileData(data);
}

void UploadMesh(Mesh *mesh, bool dynamic){
    if(mesh->colors == NULL){
        mesh->colors = (uint8_t*)RL_CALLOC(mesh->vertexCount, sizeof(RGBA8Color));
        for(size_t i = 0;i < mesh->vertexCount * 4;i++){
            mesh->colors[i] = 255;
        }
    }
    if(mesh->vbos == NULL){
        
        mesh->vbos = (DescribedBuffer**)RL_CALLOC(MAX_MESH_VERTEX_BUFFERS, sizeof(DescribedBuffer*));
        mesh->vbos[0] = GenVertexBuffer(mesh->vertices , mesh->vertexCount * sizeof(float  ) * 3);
        mesh->vbos[1] = GenVertexBuffer(mesh->texcoords, mesh->vertexCount * sizeof(float  ) * 2);
        mesh->vbos[2] = GenVertexBuffer(mesh->normals  , mesh->vertexCount * sizeof(float  ) * 3);
        mesh->vbos[3] = GenVertexBuffer(mesh->colors   , mesh->vertexCount * sizeof(uint8_t) * 4);
        if(mesh->boneWeights){
            mesh->vbos[4] = GenVertexBuffer(mesh->boneWeights, mesh->vertexCount * sizeof(float) * 4);
        }
        if(mesh->boneIds){ //TODO: Maybe change this to uint32_t
            mesh->vbos[5] = GenVertexBuffer(mesh->boneIds, mesh->vertexCount * sizeof(float) * 4);
        }

        if(mesh->indices){
            mesh->ibo = GenIndexBuffer(mesh->indices, mesh->triangleCount * 3 * sizeof(uint32_t));
        }
        mesh->vao = LoadVertexArray();
        VertexAttribPointer(mesh->vao, mesh->vbos[0], 0, WGPUVertexFormat_Float32x3, sizeof(float) * 0, WGPUVertexStepMode_Vertex);
        EnableVertexAttribArray(mesh->vao, 0);
        VertexAttribPointer(mesh->vao, mesh->vbos[1], 1, WGPUVertexFormat_Float32x2, sizeof(float) * 0, WGPUVertexStepMode_Vertex);
        EnableVertexAttribArray(mesh->vao, 1);
        VertexAttribPointer(mesh->vao, mesh->vbos[2], 2, WGPUVertexFormat_Float32x3, sizeof(float) * 0, WGPUVertexStepMode_Vertex);
        EnableVertexAttribArray(mesh->vao, 2);
        VertexAttribPointer(mesh->vao, mesh->vbos[3], 3, WGPUVertexFormat_Unorm8x4, sizeof(float) * 0, WGPUVertexStepMode_Vertex);
        EnableVertexAttribArray(mesh->vao, 3);
        if(mesh->boneWeights){
            VertexAttribPointer(mesh->vao, mesh->vbos[4], 4, WGPUVertexFormat_Float32x4, sizeof(float) * 0, WGPUVertexStepMode_Vertex);
            EnableVertexAttribArray(mesh->vao, 4);
        }
        if(mesh->boneIds){
            VertexAttribPointer(mesh->vao, mesh->vbos[5], 5, mesh->boneIDFormat, sizeof(float) * 0, WGPUVertexStepMode_Vertex);
            EnableVertexAttribArray(mesh->vao, 5);
        }
        if(mesh->boneMatrices && mesh->boneCount){
            mesh->boneMatrixBuffer = GenStorageBuffer(mesh->boneMatrices, sizeof(Matrix) * mesh->boneCount);
        }
    }
    else{
        BufferData(mesh->vbos[0], mesh->vertices , mesh->vertexCount * sizeof(float  ) * 3);
        BufferData(mesh->vbos[1], mesh->texcoords, mesh->vertexCount * sizeof(float  ) * 2);
        BufferData(mesh->vbos[2], mesh->normals  , mesh->vertexCount * sizeof(float  ) * 3);
        BufferData(mesh->vbos[3], mesh->colors   , mesh->vertexCount * sizeof(uint8_t) * 4);
        if(mesh->indices){
            BufferData(mesh->ibo, mesh->indices, mesh->triangleCount * 3 * sizeof(uint32_t));
        }
    }
}
DescribedBuffer* trfBuffer = NULL;
RGAPI void DrawMeshInstanced(Mesh mesh, Material material, const Matrix* transforms, int instances){
    if(trfBuffer && trfBuffer->buffer){
        BufferData(trfBuffer, transforms, instances * sizeof(Matrix));
    }else{
        trfBuffer = GenStorageBuffer(transforms, instances * sizeof(Matrix));
    }

    SetShaderStorageBuffer(GetActiveShader(), GetUniformLocation(GetActiveShader(), RL_DEFAULT_SHADER_UNIFORM_NAME_INSTANCE_TX), trfBuffer);
    SetTexture(GetUniformLocation(GetActiveShader(), RL_DEFAULT_SHADER_SAMPLER2D_NAME_TEXTURE0), material.maps[MATERIAL_MAP_DIFFUSE].texture);
    BindShaderVertexArray(GetActiveShader(), mesh.vao);
    if(mesh.ibo){
        DrawArraysIndexedInstanced(RL_TRIANGLES, *mesh.ibo, mesh.triangleCount * 3, instances);
    }else{
        DrawArraysInstanced(RL_TRIANGLES, mesh.vertexCount, instances);
    }
    //wgpuBufferRelease(trfBuffer.buffer);
}
RGAPI Model LoadModelFromMesh(Mesh mesh)
{
    Model model = { 0 };

    model.transform = MatrixIdentity();

    model.meshCount = 1;
    model.meshes = (Mesh *)RL_CALLOC(model.meshCount, sizeof(Mesh));
    model.meshes[0] = mesh;

    model.materialCount = 1;
    model.materials = (Material *)RL_CALLOC(model.materialCount, sizeof(Material));
    model.materials[0] = LoadMaterialDefault();

    model.meshMaterial = (int *)RL_CALLOC(model.meshCount, sizeof(int));
    model.meshMaterial[0] = 0;  // First material index

    return model;
}

RGAPI void DrawModelEx(Model model, Vector3 position, Vector3 rotationAxis, float rotationAngle, Vector3 scale, Color tint)
{
    // Calculate transformation matrix from function parameters
    // Get transform matrix (rotation -> scale -> translation)
    Matrix matScale = MatrixScale(scale.x, scale.y, scale.z);
    Matrix matRotation = MatrixRotate(rotationAxis, rotationAngle);
    Matrix matTranslation = MatrixTranslate(position.x, position.y, position.z);

    Matrix matTransform = MatrixMultiplySwap(MatrixMultiplySwap(matScale, matRotation), matTranslation);

    // Combine model transformation matrix (model.transform) with matrix generated by function parameters (matTransform)
    model.transform = matTransform;

    for (int i = 0; i < model.meshCount; i++)
    {
        Color color = model.materials[model.meshMaterial[i]].maps[MATERIAL_MAP_DIFFUSE].color;

        Color colorTint = WHITE;
        colorTint.r = (unsigned char)(((int)color.r*(int)tint.r)/255);
        colorTint.g = (unsigned char)(((int)color.g*(int)tint.g)/255);
        colorTint.b = (unsigned char)(((int)color.b*(int)tint.b)/255);
        colorTint.a = (unsigned char)(((int)color.a*(int)tint.a)/255);

        model.materials[model.meshMaterial[i]].maps[MATERIAL_MAP_DIFFUSE].color = colorTint;
        DrawMesh(model.meshes[i], model.materials[model.meshMaterial[i]], model.transform);
        model.materials[model.meshMaterial[i]].maps[MATERIAL_MAP_DIFFUSE].color = color;
    }
}

RGAPI void DrawMesh(Mesh mesh, Material material, Matrix transform){
    SetStorageBufferData(3, &transform, sizeof(Matrix));
    SetTexture(GetUniformLocation(GetActiveShader(), RL_DEFAULT_SHADER_SAMPLER2D_NAME_TEXTURE0), material.maps[MATERIAL_MAP_DIFFUSE].texture);
    BindShaderVertexArray(GetActiveShader(), mesh.vao);
    if(mesh.ibo){
        DrawArraysIndexed(RL_TRIANGLES, *mesh.ibo, mesh.triangleCount * 3);
    }else{
        DrawArrays(RL_TRIANGLES, mesh.vertexCount);
    }
}
void ProcessMaterialsOBJ(Material *materials, tinyobj_material_t *mats, int materialCount, const char* directory){
    // Init model mats
    for (int m = 0; m < materialCount; m++){
        // Init material to default
        // NOTE: Uses default shader, which only supports MATERIAL_MAP_DIFFUSE
        materials[m] = LoadMaterialDefault();

        if (mats == NULL) continue;

        // Get default texture, in case no texture is defined
        // NOTE: rlgl default texture is a 1x1 pixel UNCOMPRESSED_R8G8B8A8
        materials[m].maps[MATERIAL_MAP_DIFFUSE].texture = GetDefaultTexture();

        if (mats[m].diffuse_texname != NULL) materials[m].maps[MATERIAL_MAP_DIFFUSE].texture = LoadTexture(mats[m].diffuse_texname);  //char *diffuse_texname; // map_Kd
        else materials[m].maps[MATERIAL_MAP_DIFFUSE].color = CLITERAL(Color){ (unsigned char)(mats[m].diffuse[0]*255.0f), (unsigned char)(mats[m].diffuse[1]*255.0f), (unsigned char)(mats[m].diffuse[2]*255.0f), 255 }; //float diffuse[3];
        materials[m].maps[MATERIAL_MAP_DIFFUSE].value = 0.0f;

        if (mats[m].specular_texname != NULL) materials[m].maps[MATERIAL_MAP_SPECULAR].texture = LoadTexture(mats[m].specular_texname);  //char *specular_texname; // map_Ks
        materials[m].maps[MATERIAL_MAP_SPECULAR].color = CLITERAL(Color){ (unsigned char)(mats[m].specular[0]*255.0f), (unsigned char)(mats[m].specular[1]*255.0f), (unsigned char)(mats[m].specular[2]*255.0f), 255 }; //float specular[3];
        materials[m].maps[MATERIAL_MAP_SPECULAR].value = 0.0f;

        if (mats[m].bump_texname != NULL) materials[m].maps[MATERIAL_MAP_NORMAL].texture = LoadTexture(mats[m].bump_texname);  //char *bump_texname; // map_bump, bump
        materials[m].maps[MATERIAL_MAP_NORMAL].color = WHITE;
        materials[m].maps[MATERIAL_MAP_NORMAL].value = mats[m].shininess;

        materials[m].maps[MATERIAL_MAP_EMISSION].color = CLITERAL(Color){ (unsigned char)(mats[m].emission[0]*255.0f), (unsigned char)(mats[m].emission[1]*255.0f), (unsigned char)(mats[m].emission[2]*255.0f), 255 }; //float emission[3];

        if (mats[m].displacement_texname != NULL) materials[m].maps[MATERIAL_MAP_HEIGHT].texture = LoadTexture(mats[m].displacement_texname);  //char *displacement_texname; // disp
    }
}


Model LoadOBJ(const char *fileName){
    tinyobj_attrib_t objAttributes zeroinit;
    tinyobj_shape_t *objShapes = NULL;
    unsigned int objShapeCount = 0;

    tinyobj_material_t *objMaterials = NULL;
    unsigned int objMaterialCount = 0;

    Model model zeroinit;
    model.transform = MatrixIdentity();

    char *fileText = LoadFileText(fileName);

    if (fileText == NULL){
        TRACELOG(LOG_ERROR, "MODEL: [%s] Unable to read obj file", fileName);
        return model;
    }

    
    //strcpy(currentDir, GetWorkingDirectory()); // Save current working directory
    const char *objDir = GetDirectoryPath(fileName); // Switch to OBJ directory for material path correctness
    //if (CHDIR(workingDir) != 0) TRACELOG(LOG_WARNING, "MODEL: [%s] Failed to change working directory", workingDir);

    unsigned int dataSize = (unsigned int)strlen(fileText);

    unsigned int flags = TINYOBJ_FLAG_TRIANGULATE;
    int ret = tinyobj_parse_obj(&objAttributes, &objShapes, &objShapeCount, &objMaterials, &objMaterialCount, fileText, dataSize, flags);

    if (ret != TINYOBJ_SUCCESS)
    {
        TRACELOG(LOG_ERROR, "MODEL: Unable to read obj data %s", fileName);
        return model;
    }

    UnloadFileText(fileText);

    unsigned int faceVertIndex = 0;
    unsigned int nextShape = 1;
    int lastMaterial = -1;
    unsigned int meshIndex = 0;

    // Count meshes
    unsigned int nextShapeEnd = objAttributes.num_face_num_verts;

    // See how many verts till the next shape
    if (objShapeCount > 1) nextShapeEnd = objShapes[nextShape].face_offset;

    // Walk all the faces
    for (unsigned int faceId = 0; faceId < objAttributes.num_faces; faceId++)
    {
        if (faceId >= nextShapeEnd)
        {
            // Try to find the last vert in the next shape
            nextShape++;
            if (nextShape < objShapeCount) nextShapeEnd = objShapes[nextShape].face_offset;
            else nextShapeEnd = objAttributes.num_face_num_verts; // This is actually the total number of face verts in the file, not faces
            meshIndex++;
        }
        else if ((lastMaterial != -1) && (objAttributes.material_ids[faceId] != lastMaterial))
        {
            meshIndex++; // If this is a new material, we need to allocate a new mesh
        }

        lastMaterial = objAttributes.material_ids[faceId];
        faceVertIndex += objAttributes.face_num_verts[faceId];
    }

    // Allocate the base meshes and materials
    model.meshCount = meshIndex + 1;
    model.meshes = (Mesh *)calloc(model.meshCount, sizeof(Mesh));

    if (objMaterialCount > 0)
    {
        model.materialCount = objMaterialCount;
        model.materials = (Material *)calloc(objMaterialCount, sizeof(Material));
    }
    else // We must allocate at least one material
    {
        model.materialCount = 1;
        model.materials = (Material *)calloc(1, sizeof(Material));
    }

    model.meshMaterial = (int *)calloc(model.meshCount, sizeof(int));

    // See how many verts are in each mesh
    unsigned int *localMeshVertexCounts = (unsigned int *)calloc(model.meshCount, sizeof(unsigned int));

    faceVertIndex = 0;
    nextShapeEnd = objAttributes.num_face_num_verts;
    lastMaterial = -1;
    meshIndex = 0;
    unsigned int localMeshVertexCount = 0;

    nextShape = 1;
    if (objShapeCount > 1) nextShapeEnd = objShapes[nextShape].face_offset;

    // Walk all the faces
    for (unsigned int faceId = 0; faceId < objAttributes.num_faces; faceId++)
    {
        bool newMesh = false; // Do we need a new mesh?
        if (faceId >= nextShapeEnd)
        {
            // Try to find the last vert in the next shape
            nextShape++;
            if (nextShape < objShapeCount) nextShapeEnd = objShapes[nextShape].face_offset;
            else nextShapeEnd = objAttributes.num_face_num_verts; // this is actually the total number of face verts in the file, not faces

            newMesh = true;
        }
        else if ((lastMaterial != -1) && (objAttributes.material_ids[faceId] != lastMaterial))
        {
            newMesh = true;
        }

        lastMaterial = objAttributes.material_ids[faceId];

        if (newMesh)
        {
            localMeshVertexCounts[meshIndex] = localMeshVertexCount;

            localMeshVertexCount = 0;
            meshIndex++;
        }

        faceVertIndex += objAttributes.face_num_verts[faceId];
        localMeshVertexCount += objAttributes.face_num_verts[faceId];
    }

    localMeshVertexCounts[meshIndex] = localMeshVertexCount;

    for (int i = 0; i < model.meshCount; i++)
    {
        // Allocate the buffers for each mesh
        unsigned int vertexCount = localMeshVertexCounts[i];

            model.meshes[i].vertexCount = vertexCount;
            model.meshes[i].triangleCount = vertexCount/3;

            model.meshes[i].vertices  = (float *)   RL_CALLOC(vertexCount * 3, sizeof(float  ));
            model.meshes[i].normals   = (float *)   RL_CALLOC(vertexCount * 3, sizeof(float  ));
            model.meshes[i].texcoords = (float *)   RL_CALLOC(vertexCount * 2, sizeof(float  ));
            model.meshes[i].colors    = (uint8_t *) RL_CALLOC(vertexCount * 4, sizeof(uint8_t));
        }

    free(localMeshVertexCounts);
    localMeshVertexCounts = NULL;

    // Fill meshes
    faceVertIndex = 0;

    nextShapeEnd = objAttributes.num_face_num_verts;

    // See how many verts till the next shape
    nextShape = 1;
    if (objShapeCount > 1) nextShapeEnd = objShapes[nextShape].face_offset;
    lastMaterial = -1;
    meshIndex = 0;
    localMeshVertexCount = 0;

    // Walk all the faces
    for (unsigned int faceId = 0; faceId < objAttributes.num_faces; faceId++)
    {
        bool newMesh = false; // Do we need a new mesh?
        if (faceId >= nextShapeEnd)
        {
            // Try to find the last vert in the next shape
            nextShape++;
            if (nextShape < objShapeCount) nextShapeEnd = objShapes[nextShape].face_offset;
            else nextShapeEnd = objAttributes.num_face_num_verts; // This is actually the total number of face verts in the file, not faces
            newMesh = true;
        }

        // If this is a new material, we need to allocate a new mesh
        if (lastMaterial != -1 && objAttributes.material_ids[faceId] != lastMaterial) newMesh = true;
        lastMaterial = objAttributes.material_ids[faceId];

        if (newMesh)
        {
            localMeshVertexCount = 0;
            meshIndex++;
        }

        int matId = 0;
        if ((lastMaterial >= 0) && (lastMaterial < (int)objMaterialCount)) matId = lastMaterial;

        model.meshMaterial[meshIndex] = matId;

        for (int f = 0; f < objAttributes.face_num_verts[faceId]; f++)
        {
            int vertIndex = objAttributes.faces[faceVertIndex].v_idx;
            int normalIndex = objAttributes.faces[faceVertIndex].vn_idx;
            int texcordIndex = objAttributes.faces[faceVertIndex].vt_idx;

            for (int i = 0; i < 3; i++) model.meshes[meshIndex].vertices[localMeshVertexCount*3 + i] = objAttributes.vertices[vertIndex*3 + i];

            for (int i = 0; i < 3; i++) model.meshes[meshIndex].normals[localMeshVertexCount*3 + i] = objAttributes.normals[normalIndex*3 + i];

            for (int i = 0; i < 2; i++) model.meshes[meshIndex].texcoords[localMeshVertexCount*2 + i] = objAttributes.texcoords[texcordIndex*2 + i];

            model.meshes[meshIndex].texcoords[localMeshVertexCount*2 + 1] = 1.0f - model.meshes[meshIndex].texcoords[localMeshVertexCount*2 + 1];

            for (int i = 0; i < 4; i++) model.meshes[meshIndex].colors[localMeshVertexCount * 4 + i] = (uint8_t)255;

            faceVertIndex++;
            localMeshVertexCount++;
        }
    }

    if (objMaterialCount > 0) ProcessMaterialsOBJ(model.materials, objMaterials, objMaterialCount, objDir);
    else model.materials[0] = LoadMaterialDefault(); // Set default material for the mesh

    tinyobj_attrib_free(&objAttributes);
    tinyobj_shapes_free(objShapes, objShapeCount);
    tinyobj_materials_free(objMaterials, objMaterialCount);

    for (int i = 0; i < model.meshCount; i++) UploadMesh(model.meshes + i, true);

    // Restore current working directory
    //if (CHDIR(currentDir) != 0)
    //{
    //    TRACELOG(LOG_WARNING, "MODEL: [%s] Failed to change working directory", currentDir);
    //}

    return model;
}
static Image LoadImageFromCgltfImage(cgltf_image *cgltfImage, const char *texPath)
{
    Image image zeroinit;

    if (cgltfImage == NULL) return image;

    if (cgltfImage->uri != NULL)     // Check if image data is provided as an uri (base64 or path)
    {
        if ((strlen(cgltfImage->uri) > 5) &&
            (cgltfImage->uri[0] == 'd') &&
            (cgltfImage->uri[1] == 'a') &&
            (cgltfImage->uri[2] == 't') &&
            (cgltfImage->uri[3] == 'a') &&
            (cgltfImage->uri[4] == ':'))     // Check if image is provided as base64 text data
        {
            // Data URI Format: data:<mediatype>;base64,<data>

            // Find the comma
            int i = 0;
            while ((cgltfImage->uri[i] != ',') && (cgltfImage->uri[i] != 0)) i++;

            if (cgltfImage->uri[i] == 0) TRACELOG(LOG_WARNING, "IMAGE: glTF data URI is not a valid image");
            else
            {
                int base64Size = (int)strlen(cgltfImage->uri + i + 1);
                while (cgltfImage->uri[i + base64Size] == '=') base64Size--;    // Ignore optional paddings
                int numberOfEncodedBits = base64Size*6 - (base64Size*6) % 8 ;   // Encoded bits minus extra bits, so it becomes a multiple of 8 bits
                int outSize = numberOfEncodedBits/8 ;                           // Actual encoded bytes
                void *data = NULL;

                cgltf_options options = {0};
                options.file.read = LoadFileGLTFCallback;
                options.file.release = ReleaseFileGLTFCallback;
                cgltf_result result = cgltf_load_buffer_base64(&options, outSize, cgltfImage->uri + i + 1, &data);

                if (result == cgltf_result_success)
                {
                    image = LoadImageFromMemory(".png", (unsigned char *)data, outSize);
                    RL_FREE(data);
                }
            }
        }
        else     // Check if image is provided as image path
        {
            image = LoadImage(TextFormat("%s/%s", texPath, cgltfImage->uri));
        }
    }
    else if (cgltfImage->buffer_view != NULL && cgltfImage->buffer_view->buffer->data != NULL)    // Check if image is provided as data buffer
    {
        unsigned char *data = (unsigned char*) RL_MALLOC(cgltfImage->buffer_view->size);
        int offset = (int)cgltfImage->buffer_view->offset;
        int stride = (int)cgltfImage->buffer_view->stride? (int)cgltfImage->buffer_view->stride : 1;

        // Copy buffer data to memory for loading
        for (unsigned int i = 0; i < cgltfImage->buffer_view->size; i++)
        {
            data[i] = ((unsigned char *)cgltfImage->buffer_view->buffer->data)[offset];
            offset += stride;
        }

        // Check mime_type for image: (cgltfImage->mime_type == "image/png")
        // NOTE: Detected that some models define mime_type as "image\\/png"
        if ((strcmp(cgltfImage->mime_type, "image\\/png") == 0) ||
            (strcmp(cgltfImage->mime_type, "image/png") == 0)) image = LoadImageFromMemory(".png", data, (int)cgltfImage->buffer_view->size);
        else if ((strcmp(cgltfImage->mime_type, "image\\/jpeg") == 0) ||
                 (strcmp(cgltfImage->mime_type, "image/jpeg") == 0)) image = LoadImageFromMemory(".jpg", data, (int)cgltfImage->buffer_view->size);
        else TRACELOG(LOG_WARNING, "MODEL: glTF image data MIME type not recognized", TextFormat("%s/%s", texPath, cgltfImage->uri));

        RL_FREE(data);
    }

    return image;
}

Model LoadGLTFFromMemory(const void* fileData, size_t size){
    Model ret zeroinit;
    cgltf_options options zeroinit;
    cgltf_data* data = NULL;
    cgltf_result result = cgltf_parse(&options, fileData, size, &data);
    if (result == cgltf_result_success){
    	/* TODO make awesome stuff */
        
    	cgltf_free(data);
    }
    else{
        TRACELOG(LOG_WARNING, "GLTF could not be parsed");
    }
    return ret;
}
static BoneInfo *LoadBoneInfoGLTF(cgltf_skin skin, int *boneCount)
{
    *boneCount = (int)skin.joints_count;
    BoneInfo *bones = (BoneInfo*)RL_MALLOC(skin.joints_count*sizeof(BoneInfo));

    for (unsigned int i = 0; i < skin.joints_count; i++)
    {
        cgltf_node node = *skin.joints[i];
        if (node.name != NULL)
        {
            strncpy(bones[i].name, node.name, sizeof(bones[i].name));
            bones[i].name[sizeof(bones[i].name) - 1] = '\0';
        }

        // Find parent bone index
        int parentIndex = -1;

        for (unsigned int j = 0; j < skin.joints_count; j++)
        {
            if (skin.joints[j] == node.parent)
            {
                parentIndex = (int)j;
                break;
            }
        }

        bones[i].parent = parentIndex;
    }

    return bones;
}
void UnloadMaterial(Material mat){
    RL_FREE(mat.maps);
}
Material LoadMaterialDefault(void){
    Material material zeroinit;
    material.maps = (MaterialMap*) RL_CALLOC(MAX_MATERIAL_MAPS, sizeof(MaterialMap));

    // Using default pipeline
    material.shader = DefaultShader();

    // Using default texture (1x1 pixel, WGPuTexture, 1 mipmap)
    material.maps[MATERIAL_MAP_DIFFUSE].texture = GetDefaultTexture();
    //material.maps[MATERIAL_MAP_NORMAL].texture;         // NOTE: By default, not set
    //material.maps[MATERIAL_MAP_SPECULAR].texture;       // NOTE: By default, not set

    material.maps[MATERIAL_MAP_DIFFUSE].color = WHITE;    // Diffuse color
    material.maps[MATERIAL_MAP_SPECULAR].color = WHITE;   // Specular color

    return material;
}





static bool GetPoseAtTimeGLTF(cgltf_interpolation_type interpolationType, cgltf_accessor *input, cgltf_accessor *output, float time, void *data)
{
    if (interpolationType >= cgltf_interpolation_type_max_enum) return false;

    // Input and output should have the same count
    float tstart = 0.0f;
    float tend = 0.0f;
    int keyframe = 0;       // Defaults to first pose

    for (int i = 0; i < (int)input->count - 1; i++)
    {
        cgltf_bool r1 = cgltf_accessor_read_float(input, i, &tstart, 1);
        if (!r1) return false;

        cgltf_bool r2 = cgltf_accessor_read_float(input, i + 1, &tend, 1);
        if (!r2) return false;

        if ((tstart <= time) && (time < tend))
        {
            keyframe = i;
            break;
        }
    }

    // Constant animation, no need to interpolate
    if (FloatEquals(tend, tstart)) return true;

    float duration = fmaxf((tend - tstart), EPSILON);
    float t = (time - tstart)/duration;
    t = (t < 0.0f)? 0.0f : t;
    t = (t > 1.0f)? 1.0f : t;

    if (output->component_type != cgltf_component_type_r_32f) return false;

    if (output->type == cgltf_type_vec3)
    {
        switch (interpolationType)
        {
            case cgltf_interpolation_type_step:
            {
                float tmp[3] = { 0.0f };
                cgltf_accessor_read_float(output, keyframe, tmp, 3);
                Vector3 v1 = {tmp[0], tmp[1], tmp[2]};
                Vector3 *r = (Vector3*)data;

                *r = v1;
            } break;
            case cgltf_interpolation_type_linear:
            {
                float tmp[3] = { 0.0f };
                cgltf_accessor_read_float(output, keyframe, tmp, 3);
                Vector3 v1 = {tmp[0], tmp[1], tmp[2]};
                cgltf_accessor_read_float(output, keyframe+1, tmp, 3);
                Vector3 v2 = {tmp[0], tmp[1], tmp[2]};
                Vector3 *r = (Vector3*)data;

                *r = Vector3Lerp(v1, v2, t);
            } break;
            case cgltf_interpolation_type_cubic_spline:
            {
                float tmp[3] = { 0.0f };
                cgltf_accessor_read_float(output, 3*keyframe+1, tmp, 3);
                Vector3 v1 = {tmp[0], tmp[1], tmp[2]};
                cgltf_accessor_read_float(output, 3*keyframe+2, tmp, 3);
                Vector3 tangent1 = {tmp[0], tmp[1], tmp[2]};
                cgltf_accessor_read_float(output, 3*(keyframe+1)+1, tmp, 3);
                Vector3 v2 = {tmp[0], tmp[1], tmp[2]};
                cgltf_accessor_read_float(output, 3*(keyframe+1), tmp, 3);
                Vector3 tangent2 = {tmp[0], tmp[1], tmp[2]};
                Vector3 *r = (Vector3*)data;

                *r = Vector3CubicHermite(v1, tangent1, v2, tangent2, t);
            } break;
            default: break;
        }
    }
    else if (output->type == cgltf_type_vec4)
    {
        // Only v4 is for rotations, so we know it's a quaternion
        switch (interpolationType)
        {
            case cgltf_interpolation_type_step:
            {
                float tmp[4] = { 0.0f };
                cgltf_accessor_read_float(output, keyframe, tmp, 4);
                Vector4 v1 = {tmp[0], tmp[1], tmp[2], tmp[3]};
                Vector4 *r = (Vector4*)data;

                *r = v1;
            } break;
            case cgltf_interpolation_type_linear:
            {
                float tmp[4] = { 0.0f };
                cgltf_accessor_read_float(output, keyframe, tmp, 4);
                Vector4 v1 = {tmp[0], tmp[1], tmp[2], tmp[3]};
                cgltf_accessor_read_float(output, keyframe+1, tmp, 4);
                Vector4 v2 = {tmp[0], tmp[1], tmp[2], tmp[3]};
                Vector4 *r = (Vector4*)data;

                *r = QuaternionSlerp(v1, v2, t);
            } break;
            case cgltf_interpolation_type_cubic_spline:
            {
                float tmp[4] = { 0.0f };
                cgltf_accessor_read_float(output, 3*keyframe+1, tmp, 4);
                Vector4 v1 = {tmp[0], tmp[1], tmp[2], tmp[3]};
                cgltf_accessor_read_float(output, 3*keyframe+2, tmp, 4);
                Vector4 outTangent1 = {tmp[0], tmp[1], tmp[2], 0.0f};
                cgltf_accessor_read_float(output, 3*(keyframe+1)+1, tmp, 4);
                Vector4 v2 = {tmp[0], tmp[1], tmp[2], tmp[3]};
                cgltf_accessor_read_float(output, 3*(keyframe+1), tmp, 4);
                Vector4 inTangent2 = {tmp[0], tmp[1], tmp[2], 0.0f};
                Vector4 *r = (Vector4*)data;

                v1 = QuaternionNormalize(v1);
                v2 = QuaternionNormalize(v2);

                if (Vector4DotProduct(v1, v2) < 0.0f)
                {
                    v2 = Vector4Negate(v2);
                }
                
                outTangent1 = Vector4Scale(outTangent1, duration);
                inTangent2 = Vector4Scale(inTangent2, duration);

                *r = QuaternionCubicHermiteSpline(v1, outTangent1, v2, inTangent2, t);
            } break;
            default: break;
        }
    }

    return true;
}




static void BuildPoseFromParentJoints(BoneInfo *bones, int boneCount, Transform *transforms)
{
    for (int i = 0; i < boneCount; i++)
    {
        if (bones[i].parent >= 0)
        {
            if (bones[i].parent > i)
            {
                TRACELOG(LOG_WARNING, "Assumes bones are toplogically sorted, but bone %d has parent %d. Skipping.", i, bones[i].parent);
                continue;
            }
            transforms[i].rotation = QuaternionMultiply(transforms[bones[i].parent].rotation, transforms[i].rotation);
            transforms[i].translation = Vector3RotateByQuaternion(transforms[i].translation, transforms[bones[i].parent].rotation);
            transforms[i].translation = Vector3Add(transforms[i].translation, transforms[bones[i].parent].translation);
            transforms[i].scale = Vector3Multiply(transforms[i].scale, transforms[bones[i].parent].scale);
        }
    }
}




#define GLTF_ANIMDELAY 17

ModelAnimation *LoadModelAnimationsGLTF(const char *fileName, int *animCount){
    // glTF file loading
    size_t dataSize = 0;
    unsigned char *fileData = (unsigned char*)LoadFileData(fileName, &dataSize);

    ModelAnimation *animations = NULL;

    // glTF data loading
    cgltf_options options zeroinit;
    options.file.read = LoadFileGLTFCallback;
    options.file.release = ReleaseFileGLTFCallback;
    cgltf_data *data = NULL;
    cgltf_result result = cgltf_parse(&options, fileData, dataSize, &data);

    if (result != cgltf_result_success)
    {
        TRACELOG(LOG_WARNING, "MODEL: [%s] Failed to load glTF data", fileName);
        *animCount = 0;
        return NULL;
    }

    result = cgltf_load_buffers(&options, data, fileName);
    if (result != cgltf_result_success) TRACELOG(LOG_INFO, "MODEL: [%s] Failed to load animation buffers", fileName);

    if (result == cgltf_result_success)
    {
        if (data->skins_count > 0)
        {
            cgltf_skin skin = data->skins[0];
            *animCount = (int)data->animations_count;
            animations = (ModelAnimation*)RL_MALLOC(data->animations_count * sizeof(ModelAnimation));

            for (unsigned int i = 0; i < data->animations_count; i++)
            {
                animations[i].bones = LoadBoneInfoGLTF(skin, &animations[i].boneCount);

                cgltf_animation animData = data->animations[i];

                struct Channels {
                    cgltf_animation_channel *translate;
                    cgltf_animation_channel *rotate;
                    cgltf_animation_channel *scale;
                    cgltf_interpolation_type interpolationType;
                };

                struct Channels *boneChannels = (struct Channels *)RL_CALLOC(animations[i].boneCount, sizeof(struct Channels));
                float animDuration = 0.0f;

                for (unsigned int j = 0; j < animData.channels_count; j++)
                {
                    cgltf_animation_channel channel = animData.channels[j];
                    int boneIndex = -1;

                    for (unsigned int k = 0; k < skin.joints_count; k++)
                    {
                        if (animData.channels[j].target_node == skin.joints[k])
                        {
                            boneIndex = k;
                            break;
                        }
                    }

                    if (boneIndex == -1)
                    {
                        // Animation channel for a node not in the armature
                        continue;
                    }

                    boneChannels[boneIndex].interpolationType = animData.channels[j].sampler->interpolation;

                    if (animData.channels[j].sampler->interpolation != cgltf_interpolation_type_max_enum)
                    {
                        if (channel.target_path == cgltf_animation_path_type_translation)
                        {
                            boneChannels[boneIndex].translate = &animData.channels[j];
                        }
                        else if (channel.target_path == cgltf_animation_path_type_rotation)
                        {
                            boneChannels[boneIndex].rotate = &animData.channels[j];
                        }
                        else if (channel.target_path == cgltf_animation_path_type_scale)
                        {
                            boneChannels[boneIndex].scale = &animData.channels[j];
                        }
                        else
                        {
                            TRACELOG(LOG_WARNING, "MODEL: [%s] Unsupported target_path on channel %d's sampler for animation %d. Skipping.", fileName, j, i);
                        }
                    }
                    else TRACELOG(LOG_WARNING, "MODEL: [%s] Invalid interpolation curve encountered for GLTF animation.", fileName);

                    float t = 0.0f;
                    cgltf_bool r = cgltf_accessor_read_float(channel.sampler->input, channel.sampler->input->count - 1, &t, 1);

                    if (!r)
                    {
                        TRACELOG(LOG_WARNING, "MODEL: [%s] Failed to load input time", fileName);
                        continue;
                    }

                    animDuration = (t > animDuration)? t : animDuration;
                }

                if (animData.name != NULL)
                {
                    strncpy(animations[i].name, animData.name, sizeof(animations[i].name));
                    animations[i].name[sizeof(animations[i].name) - 1] = '\0';
                }

                animations[i].frameCount = (int)(animDuration*1000.0f/GLTF_ANIMDELAY) + 1;
                animations[i].framePoses = (Transform**)RL_MALLOC(animations[i].frameCount*sizeof(Transform *));

                for (int j = 0; j < animations[i].frameCount; j++)
                {
                    animations[i].framePoses[j] = (Transform*)RL_MALLOC(animations[i].boneCount*sizeof(Transform));
                    float time = ((float) j*GLTF_ANIMDELAY)/1000.0f;

                    for (int k = 0; k < animations[i].boneCount; k++)
                    {
                        Vector3 translation = {skin.joints[k]->translation[0], skin.joints[k]->translation[1], skin.joints[k]->translation[2]};
                        Quaternion rotation = {skin.joints[k]->rotation[0], skin.joints[k]->rotation[1], skin.joints[k]->rotation[2], skin.joints[k]->rotation[3]};
                        Vector3 scale = {skin.joints[k]->scale[0], skin.joints[k]->scale[1], skin.joints[k]->scale[2]};

                        if (boneChannels[k].translate)
                        {
                            if (!GetPoseAtTimeGLTF(boneChannels[k].interpolationType, boneChannels[k].translate->sampler->input, boneChannels[k].translate->sampler->output, time, &translation))
                            {
                                TRACELOG(LOG_INFO, "MODEL: [%s] Failed to load translate pose data for bone %s", fileName, animations[i].bones[k].name);
                            }
                        }

                        if (boneChannels[k].rotate)
                        {
                            if (!GetPoseAtTimeGLTF(boneChannels[k].interpolationType, boneChannels[k].rotate->sampler->input, boneChannels[k].rotate->sampler->output, time, &rotation))
                            {
                                TRACELOG(LOG_INFO, "MODEL: [%s] Failed to load rotate pose data for bone %s", fileName, animations[i].bones[k].name);
                            }
                        }

                        if (boneChannels[k].scale)
                        {
                            if (!GetPoseAtTimeGLTF(boneChannels[k].interpolationType, boneChannels[k].scale->sampler->input, boneChannels[k].scale->sampler->output, time, &scale))
                            {
                                TRACELOG(LOG_INFO, "MODEL: [%s] Failed to load scale pose data for bone %s", fileName, animations[i].bones[k].name);
                            }
                        }

                        animations[i].framePoses[j][k] = CLITERAL(Transform){
                            .translation = translation,
                            .rotation = rotation,
                            .scale = scale
                        };
                    }

                    BuildPoseFromParentJoints(animations[i].bones, animations[i].boneCount, animations[i].framePoses[j]);
                }

                TRACELOG(LOG_INFO, "MODEL: [%s] Loaded animation: %s (%d frames, %fs)", fileName, (animData.name != NULL)? animData.name : "NULL", animations[i].frameCount, animDuration);
                RL_FREE(boneChannels);
            }
        }

        if (data->skins_count > 1)
        {
            TRACELOG(LOG_WARNING, "MODEL: [%s] expected exactly one skin to load animation data from, but found %i", fileName, data->skins_count);
        }

        cgltf_free(data);
    }
    UnloadFileData(fileData);
    return animations;
}

Model LoadGLTF(const char *fileName)
{
    /*********************************************************************************************

        Function implemented by Wilhem Barbier(@wbrbr), with modifications by Tyler Bezera(@gamerfiend)
        Transform handling implemented by Paul Melis (@paulmelis).
        Reviewed by Ramon Santamaria (@raysan5)

        FEATURES:
          - Supports .gltf and .glb files
          - Supports embedded (base64) or external textures
          - Supports PBR metallic/roughness flow, loads material textures, values and colors
                     PBR specular/glossiness flow and extended texture flows not supported
          - Supports multiple meshes per model (every primitives is loaded as a separate mesh)
          - Supports basic animations
          - Transforms, including parent-child relations, are applied on the mesh data, but the
            hierarchy is not kept (as it can't be represented).
          - Mesh instances in the glTF file (i.e. same mesh linked from multiple nodes)
            are turned into separate raylib Meshes.

        RESTRICTIONS:
          - Only triangle meshes supported
          - Vertex attribute types and formats supported:
              > Vertices (position): vec3: float
              > Normals: vec3: float
              > Texcoords: vec2: float
              > Colors: vec4: u8, u16, f32 (normalized)
              > Indices: u16, u32 (truncated to u16)
          - Scenes defined in the glTF file are ignored. All nodes in the file
            are used.

    ***********************************************************************************************/

    // Macro to simplify attributes loading code
    #define LOAD_ATTRIBUTE(accesor, numComp, srcType, dstPtr) LOAD_ATTRIBUTE_CAST(accesor, numComp, srcType, dstPtr, srcType)

    #define LOAD_ATTRIBUTE_CAST(accesor, numComp, srcType, dstPtr, dstType) \
    { \
        int n = 0; \
        srcType *buffer = (srcType *)accesor->buffer_view->buffer->data + accesor->buffer_view->offset/sizeof(srcType) + accesor->offset/sizeof(srcType); \
        for (unsigned int k = 0; k < accesor->count; k++) \
        {\
            for (int l = 0; l < numComp; l++) \
            {\
                dstPtr[numComp*k + l] = (dstType)buffer[n + l];\
            }\
            n += (int)(accesor->stride/sizeof(srcType));\
        }\
    }

    Model model zeroinit;

    // glTF file loading
    size_t dataSize = 0;
    unsigned char *fileData = (unsigned char*)LoadFileData(fileName, &dataSize);

    if (fileData == NULL) return model;

    // glTF data loading
    cgltf_options options zeroinit;
    options.file.read =    NULL;//LoadFileGLTFCallback;
    options.file.release = NULL;//ReleaseFileGLTFCallback;
    cgltf_data *data = NULL;
    cgltf_result result = cgltf_parse(&options, fileData, dataSize, &data);

    if (result == cgltf_result_success)
    {
        if (data->file_type == cgltf_file_type_glb) TRACELOG(LOG_INFO, "MODEL: [%s] Model basic data (glb) loaded successfully", fileName);
        else if (data->file_type == cgltf_file_type_gltf) TRACELOG(LOG_INFO, "MODEL: [%s] Model basic data (glTF) loaded successfully", fileName);
        else TRACELOG(LOG_WARNING, "MODEL: [%s] Model format not recognized", fileName);

        TRACELOG(LOG_INFO, "    > Meshes count: %i", data->meshes_count);
        TRACELOG(LOG_INFO, "    > Materials count: %i (+1 default)", data->materials_count);
        TRACELOG(LOG_DEBUG, "    > Buffers count: %i", data->buffers_count);
        TRACELOG(LOG_DEBUG, "    > Images count: %i", data->images_count);
        TRACELOG(LOG_DEBUG, "    > Textures count: %i", data->textures_count);

        // Force reading data buffers (fills buffer_view->buffer->data)
        // NOTE: If an uri is defined to base64 data or external path, it's automatically loaded
        result = cgltf_load_buffers(&options, data, fileName);
        if (result != cgltf_result_success) TRACELOG(LOG_INFO, "MODEL: [%s] Failed to load mesh/material buffers", fileName);

        int primitivesCount = 0;
        // NOTE: We will load every primitive in the glTF as a separate raylib Mesh.
        // Determine total number of meshes needed from the node hierarchy.
        for (unsigned int i = 0; i < data->nodes_count; i++)
        {
            cgltf_node *node = &(data->nodes[i]);
            cgltf_mesh *mesh = node->mesh;
            if (!mesh)
                continue;

            for (unsigned int p = 0; p < mesh->primitives_count; p++)
            {
                if (mesh->primitives[p].type == cgltf_primitive_type_triangles)
                    primitivesCount++;
            }
        }
        TRACELOG(LOG_DEBUG, "    > Primitives (triangles only) count based on hierarchy : %i", primitivesCount);

        // Load our model data: meshes and materials
        model.meshCount = primitivesCount;
        model.meshes = (Mesh*)RL_CALLOC(model.meshCount, sizeof(Mesh));

        // NOTE: We keep an extra slot for default material, in case some mesh requires it
        model.materialCount = (int)data->materials_count + 1;
        model.materials = (Material*)RL_CALLOC(model.materialCount, sizeof(Material));
        model.materials[0] = LoadMaterialDefault();     // Load default material (index: 0)

        // Load mesh-material indices, by default all meshes are mapped to material index: 0
        model.meshMaterial = (int*)RL_CALLOC(model.meshCount, sizeof(int));

        // Load materials data
        //----------------------------------------------------------------------------------------------------
        for (unsigned int i = 0, j = 1; i < data->materials_count; i++, j++)
        {
            model.materials[j] = LoadMaterialDefault();
            const char *texPath = GetDirectoryPath(fileName);

            // Check glTF material flow: PBR metallic/roughness flow
            // NOTE: Alternatively, materials can follow PBR specular/glossiness flow
            if (data->materials[i].has_pbr_metallic_roughness)
            {
                // Load base color texture (albedo)
                if (data->materials[i].pbr_metallic_roughness.base_color_texture.texture)
                {
                    Image imAlbedo = LoadImageFromCgltfImage(data->materials[i].pbr_metallic_roughness.base_color_texture.texture->image, texPath);
                    if (imAlbedo.data != NULL)
                    {
                        model.materials[j].maps[MATERIAL_MAP_ALBEDO].texture = LoadTextureFromImage(imAlbedo);
                        UnloadImage(imAlbedo);
                    }
                }
                // Load base color factor (tint)
                model.materials[j].maps[MATERIAL_MAP_ALBEDO].color.r = (unsigned char)(data->materials[i].pbr_metallic_roughness.base_color_factor[0]*255);
                model.materials[j].maps[MATERIAL_MAP_ALBEDO].color.g = (unsigned char)(data->materials[i].pbr_metallic_roughness.base_color_factor[1]*255);
                model.materials[j].maps[MATERIAL_MAP_ALBEDO].color.b = (unsigned char)(data->materials[i].pbr_metallic_roughness.base_color_factor[2]*255);
                model.materials[j].maps[MATERIAL_MAP_ALBEDO].color.a = (unsigned char)(data->materials[i].pbr_metallic_roughness.base_color_factor[3]*255);

                // Load metallic/roughness texture
                if (data->materials[i].pbr_metallic_roughness.metallic_roughness_texture.texture)
                {
                    Image imMetallicRoughness = LoadImageFromCgltfImage(data->materials[i].pbr_metallic_roughness.metallic_roughness_texture.texture->image, texPath);
                    if (imMetallicRoughness.data != NULL)
                    {
                        model.materials[j].maps[MATERIAL_MAP_ROUGHNESS].texture = LoadTextureFromImage(imMetallicRoughness);
                        UnloadImage(imMetallicRoughness);
                    }

                    // Load metallic/roughness material properties
                    float roughness = data->materials[i].pbr_metallic_roughness.roughness_factor;
                    model.materials[j].maps[MATERIAL_MAP_ROUGHNESS].value = roughness;

                    float metallic = data->materials[i].pbr_metallic_roughness.metallic_factor;
                    model.materials[j].maps[MATERIAL_MAP_METALNESS].value = metallic;
                }

                // Load normal texture
                if (data->materials[i].normal_texture.texture)
                {
                    Image imNormal = LoadImageFromCgltfImage(data->materials[i].normal_texture.texture->image, texPath);
                    if (imNormal.data != NULL)
                    {
                        model.materials[j].maps[MATERIAL_MAP_NORMAL].texture = LoadTextureFromImage(imNormal);
                        UnloadImage(imNormal);
                    }
                }

                // Load ambient occlusion texture
                if (data->materials[i].occlusion_texture.texture)
                {
                    Image imOcclusion = LoadImageFromCgltfImage(data->materials[i].occlusion_texture.texture->image, texPath);
                    if (imOcclusion.data != NULL)
                    {
                        model.materials[j].maps[MATERIAL_MAP_OCCLUSION].texture = LoadTextureFromImage(imOcclusion);
                        UnloadImage(imOcclusion);
                    }
                }

                // Load emissive texture
                if (data->materials[i].emissive_texture.texture)
                {
                    Image imEmissive = LoadImageFromCgltfImage(data->materials[i].emissive_texture.texture->image, texPath);
                    if (imEmissive.data != NULL)
                    {
                        model.materials[j].maps[MATERIAL_MAP_EMISSION].texture = LoadTextureFromImage(imEmissive);
                        UnloadImage(imEmissive);
                    }

                    // Load emissive color factor
                    model.materials[j].maps[MATERIAL_MAP_EMISSION].color.r = (unsigned char)(data->materials[i].emissive_factor[0]*255);
                    model.materials[j].maps[MATERIAL_MAP_EMISSION].color.g = (unsigned char)(data->materials[i].emissive_factor[1]*255);
                    model.materials[j].maps[MATERIAL_MAP_EMISSION].color.b = (unsigned char)(data->materials[i].emissive_factor[2]*255);
                    model.materials[j].maps[MATERIAL_MAP_EMISSION].color.a = 255;
                }
            }

            // Other possible materials not supported by raylib pipeline:
            // has_clearcoat, has_transmission, has_volume, has_ior, has specular, has_sheen
        }

        // Visit each node in the hierarchy and process any mesh linked from it.
        // Each primitive within a glTF node becomes a Raylib Mesh.
        // The local-to-world transform of each node is used to transform the
        // points/normals/tangents of the created Mesh(es).
        // Any glTF mesh linked from more than one Node (i.e. instancing)
        // is turned into multiple Mesh's, as each Node will have its own
        // transform applied.
        // Note: the code below disregards the scenes defined in the file, all nodes are used.
        //----------------------------------------------------------------------------------------------------
        int meshIndex = 0;
        for (unsigned int i = 0; i < data->nodes_count; i++)
        {
            cgltf_node *node = &(data->nodes[i]);

            cgltf_mesh *mesh = node->mesh;
            if (!mesh)
                continue;

            cgltf_float worldTransform[16];
            cgltf_node_transform_world(node, worldTransform);

            Matrix worldMatrix = {
                worldTransform[0], worldTransform[4], worldTransform[8], worldTransform[12],
                worldTransform[1], worldTransform[5], worldTransform[9], worldTransform[13],
                worldTransform[2], worldTransform[6], worldTransform[10], worldTransform[14],
                worldTransform[3], worldTransform[7], worldTransform[11], worldTransform[15]
            };

            Matrix worldMatrixNormals = MatrixTranspose(MatrixInvert(worldMatrix));

            for (unsigned int p = 0; p < mesh->primitives_count; p++)
            {
                // NOTE: We only support primitives defined by triangles
                // Other alternatives: points, lines, line_strip, triangle_strip
                if (mesh->primitives[p].type != cgltf_primitive_type_triangles) continue;

                // NOTE: Attributes data could be provided in several data formats (8, 8u, 16u, 32...),
                // Only some formats for each attribute type are supported, read info at the top of this function!

                for (unsigned int j = 0; j < mesh->primitives[p].attributes_count; j++)
                {
                    // Check the different attributes for every primitive
                    if (mesh->primitives[p].attributes[j].type == cgltf_attribute_type_position)      // POSITION, vec3, float
                    {
                        cgltf_accessor *attribute = mesh->primitives[p].attributes[j].data;

                        // WARNING: SPECS: POSITION accessor MUST have its min and max properties defined

                        if ((attribute->type == cgltf_type_vec3) && (attribute->component_type == cgltf_component_type_r_32f))
                        {
                            // Init raylib mesh vertices to copy glTF attribute data
                            model.meshes[meshIndex].vertexCount = (int)attribute->count;
                            model.meshes[meshIndex].vertices = (float*)RL_MALLOC(attribute->count*3*sizeof(float));

                            // Load 3 components of float data type into mesh.vertices
                            LOAD_ATTRIBUTE(attribute, 3, float, model.meshes[meshIndex].vertices)

                            // Transform the vertices
                            float *vertices = model.meshes[meshIndex].vertices;
                            for (unsigned int k = 0; k < attribute->count; k++)
                            {
                                //Vector3 vt = worldMatrix * CLITERAL(Vector3){ vertices[3*k], vertices[3*k+1], vertices[3*k+2] };
                                Vector3 vt = Vector3Transform((Vector3){ vertices[3*k], vertices[3*k+1], vertices[3*k+2] }, worldMatrix);
                                vertices[3*k] = vt.x;
                                vertices[3*k+1] = vt.y;
                                vertices[3*k+2] = vt.z;
                            }
                        }
                        else TRACELOG(LOG_WARNING, "MODEL: [%s] Vertices attribute data format not supported, use vec3 float", fileName);
                    }
                    else if (mesh->primitives[p].attributes[j].type == cgltf_attribute_type_normal)   // NORMAL, vec3, float
                    {
                        cgltf_accessor *attribute = mesh->primitives[p].attributes[j].data;

                        if ((attribute->type == cgltf_type_vec3) && (attribute->component_type == cgltf_component_type_r_32f))
                        {
                            // Init raylib mesh normals to copy glTF attribute data
                            model.meshes[meshIndex].normals = (float*)RL_MALLOC(attribute->count*3*sizeof(float));

                            // Load 3 components of float data type into mesh.normals
                            LOAD_ATTRIBUTE(attribute, 3, float, model.meshes[meshIndex].normals)

                            // Transform the normals
                            float *normals = model.meshes[meshIndex].normals;
                            for (unsigned int k = 0; k < attribute->count; k++)
                            {
                                //Vector3 nt = worldMatrixNormals * CLITERAL(Vector3){ normals[3*k], normals[3*k+1], normals[3*k+2] };
                                Vector3 nt = Vector3Transform(CLITERAL(Vector3){ normals[3*k], normals[3*k+1], normals[3*k+2] }, worldMatrixNormals);
                                normals[3*k] = nt.x;
                                normals[3*k+1] = nt.y;
                                normals[3*k+2] = nt.z;
                            }
                        }
                        else TRACELOG(LOG_WARNING, "MODEL: [%s] Normal attribute data format not supported, use vec3 float", fileName);
                    }
                    else if (mesh->primitives[p].attributes[j].type == cgltf_attribute_type_tangent)   // TANGENT, vec3, float
                    {
                        cgltf_accessor *attribute = mesh->primitives[p].attributes[j].data;

                        if ((attribute->type == cgltf_type_vec4) && (attribute->component_type == cgltf_component_type_r_32f))
                        {
                            // Init raylib mesh tangent to copy glTF attribute data
                            model.meshes[meshIndex].tangents = (float*)RL_MALLOC(attribute->count*4*sizeof(float));

                            // Load 4 components of float data type into mesh.tangents
                            LOAD_ATTRIBUTE(attribute, 4, float, model.meshes[meshIndex].tangents)

                            // Transform the tangents
                            float *tangents = model.meshes[meshIndex].tangents;
                            for (unsigned int k = 0; k < attribute->count; k++)
                            {
                                //Vector3 tt = worldMatrix * CLITERAL(Vector3){tangents[3*k], tangents[3*k+1], tangents[3*k+2]};
                                Vector3 tt =  Vector3Transform(CLITERAL(Vector3){tangents[3*k], tangents[3*k+1], tangents[3*k+2]}, worldMatrix);
                                tangents[3*k] = tt.x;
                                tangents[3*k+1] = tt.y;
                                tangents[3*k+2] = tt.z;
                            }
                        }
                        else TRACELOG(LOG_WARNING, "MODEL: [%s] Tangent attribute data format not supported, use vec4 float", fileName);
                    }
                    else if (mesh->primitives[p].attributes[j].type == cgltf_attribute_type_texcoord) // TEXCOORD_n, vec2, float/u8n/u16n
                    {
                        // Support up to 2 texture coordinates attributes
                        float *texcoordPtr = NULL;

                        cgltf_accessor *attribute = mesh->primitives[p].attributes[j].data;

                        if (attribute->type == cgltf_type_vec2)
                        {
                            if (attribute->component_type == cgltf_component_type_r_32f)  // vec2, float
                            {
                                // Init raylib mesh texcoords to copy glTF attribute data
                                texcoordPtr = (float *)RL_MALLOC(attribute->count*2*sizeof(float));

                                // Load 3 components of float data type into mesh.texcoords
                                LOAD_ATTRIBUTE(attribute, 2, float, texcoordPtr)
                            }
                            else if (attribute->component_type == cgltf_component_type_r_8u) // vec2, u8n
                            {
                                // Init raylib mesh texcoords to copy glTF attribute data
                                texcoordPtr = (float *)RL_MALLOC(attribute->count*2*sizeof(float));

                                // Load data into a temp buffer to be converted to raylib data type
                                unsigned char *temp = (unsigned char *)RL_MALLOC(attribute->count*2*sizeof(unsigned char));
                                LOAD_ATTRIBUTE(attribute, 2, unsigned char, temp)

                                // Convert data to raylib texcoord data type (float)
                                for (unsigned int t = 0; t < attribute->count*2; t++) texcoordPtr[t] = (float)temp[t]/255.0f;

                                RL_FREE(temp);
                            }
                            else if (attribute->component_type == cgltf_component_type_r_16u) // vec2, u16n
                            {
                                // Init raylib mesh texcoords to copy glTF attribute data
                                texcoordPtr = (float *)RL_MALLOC(attribute->count*2*sizeof(float));

                                // Load data into a temp buffer to be converted to raylib data type
                                unsigned short *temp = (unsigned short *)RL_MALLOC(attribute->count*2*sizeof(unsigned short));
                                LOAD_ATTRIBUTE(attribute, 2, unsigned short, temp)

                                // Convert data to raylib texcoord data type (float)
                                for (unsigned int t = 0; t < attribute->count*2; t++) texcoordPtr[t] = (float)temp[t]/65535.0f;

                                RL_FREE(temp);
                            }
                            else TRACELOG(LOG_WARNING, "MODEL: [%s] Texcoords attribute data format not supported", fileName);
                        }
                        else TRACELOG(LOG_WARNING, "MODEL: [%s] Texcoords attribute data format not supported, use vec2 float", fileName);

                        int index = mesh->primitives[p].attributes[j].index;
                        if (index == 0) model.meshes[meshIndex].texcoords = texcoordPtr;
                        else if (index == 1) model.meshes[meshIndex].texcoords2 = texcoordPtr;
                        else
                        {
                            TRACELOG(LOG_WARNING, "MODEL: [%s] No more than 2 texture coordinates attributes supported", fileName);
                            if (texcoordPtr != NULL) RL_FREE(texcoordPtr);
                        }
                    }
                    else if (mesh->primitives[p].attributes[j].type == cgltf_attribute_type_color)    // COLOR_n, vec3/vec4, float/u8n/u16n
                    {
                        cgltf_accessor *attribute = mesh->primitives[p].attributes[j].data;

                        // WARNING: SPECS: All components of each COLOR_n accessor element MUST be clamped to [0.0, 1.0] range

                        if (attribute->type == cgltf_type_vec3)  // RGB
                        {
                            if (attribute->component_type == cgltf_component_type_r_8u)
                            {
                                // Init raylib mesh color to copy glTF attribute data
                                model.meshes[meshIndex].colors = (uint8_t*)RL_MALLOC(attribute->count * 4 * sizeof(uint8_t));

                                // Load data into a temp buffer to be converted to raylib data type
                                unsigned char *temp = (unsigned char*)RL_MALLOC(attribute->count*3*sizeof(unsigned char));

                                LOAD_ATTRIBUTE(attribute, 3, unsigned char, temp)

                                // Convert data to raylib color data type (4 bytes)
                                for (unsigned int c = 0, k = 0; c < (attribute->count*4 - 3); c += 4, k += 3)
                                {
                                    model.meshes[meshIndex].colors[c] = temp[k] / 255.0f;
                                    model.meshes[meshIndex].colors[c + 1] = temp[k + 1] / 255.0f;
                                    model.meshes[meshIndex].colors[c + 2] = temp[k + 2] / 255.0f;
                                    model.meshes[meshIndex].colors[c + 3] = 1.0f;
                                }

                                RL_FREE(temp);
                            }
                            else if (attribute->component_type == cgltf_component_type_r_16u)
                            {
                                // Init raylib mesh color to copy glTF attribute data
                                model.meshes[meshIndex].colors = (uint8_t*)RL_MALLOC(attribute->count * 4 * sizeof(uint8_t));

                                // Load data into a temp buffer to be converted to raylib data type
                                unsigned short *temp = (unsigned short *)RL_MALLOC(attribute->count*3*sizeof(unsigned short));
                                LOAD_ATTRIBUTE(attribute, 3, unsigned short, temp)

                                // Convert data to raylib color data type (4 bytes)
                                for (unsigned int c = 0, k = 0; c < (attribute->count*4 - 3); c += 4, k += 3)
                                {
                                    model.meshes[meshIndex].colors[c] = (unsigned char)(((float)temp[k] / 255));
                                    model.meshes[meshIndex].colors[c + 1] = (unsigned char)(((float)temp[k + 1] / 255));
                                    model.meshes[meshIndex].colors[c + 2] = (unsigned char)(((float)temp[k + 2] / 255));
                                    model.meshes[meshIndex].colors[c + 3] = (uint8_t)(255);
                                }

                                RL_FREE(temp);
                            }
                            else if (attribute->component_type == cgltf_component_type_r_32f)
                            {
                                // Init raylib mesh color to copy glTF attribute data
                                model.meshes[meshIndex].colors = (uint8_t*)RL_MALLOC(attribute->count*4*sizeof(uint8_t));

                                // Load data into a temp buffer to be converted to raylib data type
                                float *temp = (float *)RL_MALLOC(attribute->count*3*sizeof(float));
                                LOAD_ATTRIBUTE(attribute, 3, float, temp)

                                // Convert data to raylib color data type (4 bytes)
                                for (unsigned int c = 0, k = 0; c < (attribute->count*4 - 3); c += 4, k += 3)
                                {
                                    model.meshes[meshIndex].colors[c]     = (unsigned char)(temp[k]     * 255.0f);
                                    model.meshes[meshIndex].colors[c + 1] = (unsigned char)(temp[k + 1] * 255.0f);
                                    model.meshes[meshIndex].colors[c + 2] = (unsigned char)(temp[k + 2] * 255.0f);
                                    model.meshes[meshIndex].colors[c + 3] = (uint8_t)(255);
                                }

                                RL_FREE(temp);
                            }
                            else TRACELOG(LOG_WARNING, "MODEL: [%s] Color attribute data format not supported", fileName);
                        }
                        else if (attribute->type == cgltf_type_vec4) // RGBA
                        {
                            if (attribute->component_type == cgltf_component_type_r_8u)
                            {
                                // Init raylib mesh color to copy glTF attribute data
                                model.meshes[meshIndex].colors = (uint8_t*)RL_MALLOC(attribute->count*4*sizeof(uint8_t));

                                unsigned char *temp = (unsigned char*)RL_MALLOC(attribute->count*4*sizeof(unsigned char));
                                LOAD_ATTRIBUTE(attribute, 4, unsigned char, temp)
                                for (unsigned int c = 0; c < attribute->count*4; c++) model.meshes[meshIndex].colors[c] = temp[c];

                                RL_FREE(temp);
                            }
                            else if (attribute->component_type == cgltf_component_type_r_16u)
                            {
                                // Init raylib mesh color to copy glTF attribute data
                                model.meshes[meshIndex].colors = (uint8_t*)RL_MALLOC(attribute->count*4*sizeof(uint8_t));

                                // Load data into a temp buffer to be converted to raylib data type
                                uint16_t *temp = (uint16_t*)RL_MALLOC(attribute->count*4*sizeof(uint16_t));
                                LOAD_ATTRIBUTE(attribute, 4, uint16_t, temp)

                                // Convert data to raylib color data type (4 bytes)
                                for (unsigned int c = 0; c < attribute->count*4; c++) model.meshes[meshIndex].colors[c] = (((uint8_t)(temp[c] / 255)));

                                RL_FREE(temp);
                            }
                            else if (attribute->component_type == cgltf_component_type_r_32f)
                            {
                                // Init raylib mesh color to copy glTF attribute data
                                model.meshes[meshIndex].colors = (uint8_t*)RL_MALLOC(attribute->count * 4 * sizeof(uint8_t));

                                // Load data into a temp buffer to be converted to raylib data type
                                float *temp = (float*)RL_MALLOC(attribute->count*4*sizeof(float));
                                LOAD_ATTRIBUTE(attribute, 4, float, temp)

                                // Convert data to raylib color data type (4 bytes), we expect the color data normalized
                                for (unsigned int c = 0; c < attribute->count*4; c++) model.meshes[meshIndex].colors[c] = (uint8_t)(temp[c] * 255);

                                RL_FREE(temp);
                            }
                            else TRACELOG(LOG_WARNING, "MODEL: [%s] Color attribute data format not supported", fileName);
                        }
                        else TRACELOG(LOG_WARNING, "MODEL: [%s] Color attribute data format not supported", fileName);
                    }

                    // NOTE: Attributes related to animations are processed separately
                }

                // Load primitive indices data (if provided)
                if ((mesh->primitives[p].indices != NULL) && (mesh->primitives[p].indices->buffer_view != NULL))
                {
                    cgltf_accessor *attribute = mesh->primitives[p].indices;

                    model.meshes[meshIndex].triangleCount = (int)attribute->count/3;

                    if (attribute->component_type == cgltf_component_type_r_16u)
                    {
                        // Init raylib mesh indices to copy glTF attribute data
                        model.meshes[meshIndex].indices = (uint32_t*)RL_MALLOC(attribute->count*sizeof(uint32_t));

                        // Load unsigned short data type into mesh.indices
                        LOAD_ATTRIBUTE(attribute, 1, unsigned short, model.meshes[meshIndex].indices)
                    }
                    else if (attribute->component_type == cgltf_component_type_r_8u)
                    {
                        // Init raylib mesh indices to copy glTF attribute data
                        model.meshes[meshIndex].indices = (uint32_t*)RL_MALLOC(attribute->count*sizeof(uint32_t));
                        LOAD_ATTRIBUTE_CAST(attribute, 1, unsigned char, model.meshes[meshIndex].indices, uint32_t)

                    }
                    else if (attribute->component_type == cgltf_component_type_r_32u)
                    {
                        // Init raylib mesh indices to copy glTF attribute data
                        model.meshes[meshIndex].indices = (uint32_t*)RL_MALLOC(attribute->count*sizeof(uint32_t));
                        LOAD_ATTRIBUTE_CAST(attribute, 1, unsigned int, model.meshes[meshIndex].indices, uint32_t)

                        TRACELOG(LOG_WARNING, "MODEL: [%s] Indices data converted from u32 to u16, possible loss of data", fileName);
                    }
                    else
                    {
                        TRACELOG(LOG_WARNING, "MODEL: [%s] Indices data format not supported, use u16", fileName);
                    }
                }
                else model.meshes[meshIndex].triangleCount = model.meshes[meshIndex].vertexCount/3;    // Unindexed mesh

                // Assign to the primitive mesh the corresponding material index
                // NOTE: If no material defined, mesh uses the already assigned default material (index: 0)
                for (unsigned int m = 0; m < data->materials_count; m++)
                {
                    // The primitive actually keeps the pointer to the corresponding material,
                    // raylib instead assigns to the mesh the by its index, as loaded in model.materials array
                    // To get the index, we check if material pointers match, and we assign the corresponding index,
                    // skipping index 0, the default material
                    if (&data->materials[m] == mesh->primitives[p].material)
                    {
                        model.meshMaterial[meshIndex] = m + 1;
                        break;
                    }
                }

                meshIndex++;       // Move to next mesh
            }
        }

        // Load glTF meshes animation data
        // REF: https://www.khronos.org/registry/glTF/specs/2.0/glTF-2.0.html#skins
        // REF: https://www.khronos.org/registry/glTF/specs/2.0/glTF-2.0.html#skinned-mesh-attributes
        //
        // LIMITATIONS:
        //  - Only supports 1 armature per file, and skips loading it if there are multiple armatures
        //  - Only supports linear interpolation (default method in Blender when checked "Always Sample Animations" when exporting a GLTF file)
        //  - Only supports translation/rotation/scale animation channel.path, weights not considered (i.e. morph targets)
        //----------------------------------------------------------------------------------------------------
        if (data->skins_count > 0)
        {
            cgltf_skin skin = data->skins[0];
            model.bones = LoadBoneInfoGLTF(skin, &model.boneCount);
            model.bindPose = (Transform*)RL_MALLOC(model.boneCount*sizeof(Transform));

            for (int i = 0; i < model.boneCount; i++)
            {
                cgltf_node *node = skin.joints[i];
                cgltf_float worldTransform[16];
                cgltf_node_transform_world(node, worldTransform);
                Matrix worldMatrix = {
                    worldTransform[0], worldTransform[4], worldTransform[8], worldTransform[12],
                    worldTransform[1], worldTransform[5], worldTransform[9], worldTransform[13],
                    worldTransform[2], worldTransform[6], worldTransform[10], worldTransform[14],
                    worldTransform[3], worldTransform[7], worldTransform[11], worldTransform[15]
                };
                worldMatrix = MatrixTranspose(worldMatrix);
                MatrixDecompose(worldMatrix, &(model.bindPose[i].translation), &(model.bindPose[i].rotation), &(model.bindPose[i].scale));
            }
        }
        if (data->skins_count > 1)
        {
            TRACELOG(LOG_WARNING, "MODEL: [%s] can only load one skin (armature) per model, but gltf skins_count == %i", fileName, data->skins_count);
        }

        meshIndex = 0;
        for (unsigned int i = 0; i < data->nodes_count; i++)
        {
            cgltf_node *node = &(data->nodes[i]);

            cgltf_mesh *mesh = node->mesh;
            if (!mesh)
                continue;

            for (unsigned int p = 0; p < mesh->primitives_count; p++)
            {
                // NOTE: We only support primitives defined by triangles
                if (mesh->primitives[p].type != cgltf_primitive_type_triangles) continue;

                for (unsigned int j = 0; j < mesh->primitives[p].attributes_count; j++)
                {
                    // NOTE: JOINTS_1 + WEIGHT_1 will be used for +4 joints influencing a vertex -> Not supported by raylib

                    if (mesh->primitives[p].attributes[j].type == cgltf_attribute_type_joints) // JOINTS_n (vec4: 4 bones max per vertex / u8, u16)
                    {
                        cgltf_accessor *attribute = mesh->primitives[p].attributes[j].data;

                        // NOTE: JOINTS_n can only be vec4 and u8/u16
                        // SPECS: https://registry.khronos.org/glTF/specs/2.0/glTF-2.0.html#meshes-overview

                        // WARNING: raylib only supports model.meshes[].boneIds as u8 (unsigned char),
                        // if data is provided in any other format, it is converted to supported format but
                        // it could imply data loss (a warning message is issued in that case)

                        if (attribute->type == cgltf_type_vec4)
                        {
                            if (attribute->component_type == cgltf_component_type_r_8u)
                            {
                                // Init raylib mesh boneIds to copy glTF attribute data
                                model.meshes[meshIndex].boneIds = (unsigned char*)RL_CALLOC(model.meshes[meshIndex].vertexCount*4, sizeof(unsigned char));

                                // Load attribute: vec4, u8 (unsigned char)
                                LOAD_ATTRIBUTE(attribute, 4, unsigned char, model.meshes[meshIndex].boneIds)
                                model.meshes[meshIndex].boneIDFormat = WGPUVertexFormat_Uint8x4;
                            }
                            else if (attribute->component_type == cgltf_component_type_r_16u)
                            {
                                // Init raylib mesh boneIds to copy glTF attribute data
                                //model.meshes[meshIndex].boneIds = (unsigned char*)RL_CALLOC(model.meshes[meshIndex].vertexCount*4, sizeof(unsigned char));

                                // Load data into a temp buffer to be converted to raylib data type
                                model.meshes[meshIndex].boneIds = (unsigned char*)RL_CALLOC(model.meshes[meshIndex].vertexCount*4, sizeof(unsigned short));
                                
                                LOAD_ATTRIBUTE(attribute, 4, unsigned short, model.meshes[meshIndex].boneIds)

                                model.meshes[meshIndex].boneIDFormat = WGPUVertexFormat_Uint16x4;
                            }
                            else TRACELOG(LOG_WARNING, "MODEL: [%s] Joint attribute data format not supported", fileName);
                        }
                        else TRACELOG(LOG_WARNING, "MODEL: [%s] Joint attribute data format not supported", fileName);
                    }
                    else if (mesh->primitives[p].attributes[j].type == cgltf_attribute_type_weights)  // WEIGHTS_n (vec4, u8n/u16n/f32)
                    {
                        cgltf_accessor *attribute = mesh->primitives[p].attributes[j].data;

                        if (attribute->type == cgltf_type_vec4)
                        {
                            // TODO: Support component types: u8, u16?
                            if (attribute->component_type == cgltf_component_type_r_8u)
                            {
                                // Init raylib mesh bone weight to copy glTF attribute data
                                model.meshes[meshIndex].boneWeights = (float*)RL_CALLOC(model.meshes[meshIndex].vertexCount*4, sizeof(float));

                                // Load data into a temp buffer to be converted to raylib data type
                                unsigned char *temp = (unsigned char*)RL_MALLOC(attribute->count*4*sizeof(unsigned char));
                                LOAD_ATTRIBUTE(attribute, 4, unsigned char, temp)

                                // Convert data to raylib bone weight data type (4 bytes)
                                for (unsigned int b = 0; b < attribute->count*4; b++) model.meshes[meshIndex].boneWeights[b] = (float)temp[b]/255.0f;

                                RL_FREE(temp);
                            }
                            else if (attribute->component_type == cgltf_component_type_r_16u)
                            {
                                // Init raylib mesh bone weight to copy glTF attribute data
                                model.meshes[meshIndex].boneWeights = (float*)RL_CALLOC(model.meshes[meshIndex].vertexCount*4, sizeof(float));

                                // Load data into a temp buffer to be converted to raylib data type
                                unsigned short *temp = (unsigned short*)RL_MALLOC(attribute->count*4*sizeof(unsigned short));
                                LOAD_ATTRIBUTE(attribute, 4, unsigned short, temp)

                                // Convert data to raylib bone weight data type
                                for (unsigned int b = 0; b < attribute->count*4; b++) model.meshes[meshIndex].boneWeights[b] = (float)temp[b]/65535.0f;

                                RL_FREE(temp);
                            }
                            else if (attribute->component_type == cgltf_component_type_r_32f)
                            {
                                // Init raylib mesh bone weight to copy glTF attribute data
                                model.meshes[meshIndex].boneWeights = (float*)RL_CALLOC(model.meshes[meshIndex].vertexCount*4, sizeof(float));

                                // Load 4 components of float data type into mesh.boneWeights
                                // for cgltf_attribute_type_weights we have:
                                //   - data.meshes[0] (256 vertices)
                                //   - 256 values, provided as cgltf_type_vec4 of float (4 byte per joint, stride 16)
                                LOAD_ATTRIBUTE(attribute, 4, float, model.meshes[meshIndex].boneWeights)
                            }
                            else TRACELOG(LOG_WARNING, "MODEL: [%s] Joint weight attribute data format not supported, use vec4 float", fileName);
                        }
                        else TRACELOG(LOG_WARNING, "MODEL: [%s] Joint weight attribute data format not supported, use vec4 float", fileName);
                    }
                }

                // Animated vertex data
                model.meshes[meshIndex].animVertices = (float*)RL_CALLOC(model.meshes[meshIndex].vertexCount*3, sizeof(float));
                memcpy(model.meshes[meshIndex].animVertices, model.meshes[meshIndex].vertices, model.meshes[meshIndex].vertexCount*3*sizeof(float));
                model.meshes[meshIndex].animNormals = (float*)RL_CALLOC(model.meshes[meshIndex].vertexCount*3, sizeof(float));
                if (model.meshes[meshIndex].normals != NULL)
                {
                    memcpy(model.meshes[meshIndex].animNormals, model.meshes[meshIndex].normals, model.meshes[meshIndex].vertexCount*3*sizeof(float));
                }

                // Bone Transform Matrices
                model.meshes[meshIndex].boneCount = model.boneCount;
                model.meshes[meshIndex].boneMatrices = (Matrix*)RL_CALLOC(model.meshes[meshIndex].boneCount, sizeof(Matrix));

                for (int j = 0; j < model.meshes[meshIndex].boneCount; j++)
                {
                    model.meshes[meshIndex].boneMatrices[j] = MatrixIdentity();
                }

                meshIndex++;       // Move to next mesh
            }

        }

        // Free all cgltf loaded data
        cgltf_free(data);
    }
    else TRACELOG(LOG_WARNING, "MODEL: [%s] Failed to load glTF data", fileName);

    // WARNING: cgltf requires the file pointer available while reading data
    UnloadFileData(fileData);

    return model;
}

// Update model animated bones transform matrices for a given frame
// NOTE: Updated data is not uploaded to GPU but kept at model.meshes[i].boneMatrices[boneId],
// to be uploaded to shader at drawing, in case GPU skinning is enabled
void UpdateModelAnimationBones(Model model, ModelAnimation anim, int frame)
{
    if ((anim.frameCount > 0) && (anim.bones != NULL) && (anim.framePoses != NULL))
    {
        if (frame >= anim.frameCount) frame = frame%anim.frameCount;

        // Get first mesh which have bones
        int firstMeshWithBones = -1;

        for (int i = 0; i < model.meshCount; i++)
        {
            if (model.meshes[i].boneMatrices)
            {
                if (firstMeshWithBones == -1)
                {
                    firstMeshWithBones = i;
                    break;
                }
            }
        }

        // Update all bones and boneMatrices of first mesh with bones.
        for (int boneId = 0; boneId < anim.boneCount; boneId++)
        {
            Vector3 inTranslation = model.bindPose[boneId].translation;
            Quaternion inRotation = model.bindPose[boneId].rotation;
            Vector3 inScale = model.bindPose[boneId].scale;

            Vector3 outTranslation = anim.framePoses[frame][boneId].translation;
            Quaternion outRotation = anim.framePoses[frame][boneId].rotation;
            Vector3 outScale = anim.framePoses[frame][boneId].scale;

            Quaternion invRotation = QuaternionInvert(inRotation);
            Vector3 invTranslation = Vector3RotateByQuaternion(Vector3Negate(inTranslation), invRotation);
            Vector3 invScale = Vector3Divide(CLITERAL(Vector3){ 1.0f, 1.0f, 1.0f }, inScale);

            Vector3 boneTranslation = Vector3Add(Vector3RotateByQuaternion(
                Vector3Multiply(outScale, invTranslation), outRotation), outTranslation);
            Quaternion boneRotation = QuaternionMultiply(outRotation, invRotation);
            Vector3 boneScale = Vector3Multiply(outScale, invScale);

            Matrix boneMatrix = (MatrixMultiplySwap(MatrixMultiplySwap(
                QuaternionToMatrix(boneRotation),
                MatrixTranslate(boneTranslation.x, boneTranslation.y, boneTranslation.z)),
                MatrixScale(boneScale.x, boneScale.y, boneScale.z)));

            model.meshes[firstMeshWithBones].boneMatrices[boneId] = boneMatrix;
        }

        // Update remaining meshes with bones
        // NOTE: Using deep copy because shallow copy results in double free with 'UnloadModel()'
        if (firstMeshWithBones != -1)
        {
            for (int i = firstMeshWithBones + 1; i < model.meshCount; i++)
            {
                if (model.meshes[i].boneMatrices)
                {
                    memcpy(model.meshes[i].boneMatrices,
                        model.meshes[firstMeshWithBones].boneMatrices,
                        model.meshes[i].boneCount*sizeof(model.meshes[i].boneMatrices[0]));
                }
            }
        }
    }
}

// at least 2x speed up vs the old method
// Update model animated vertex data (positions and normals) for a given frame
// NOTE: Updated data is uploaded to GPU
void UpdateModelAnimation(Model model, ModelAnimation anim, int frame)
{
    UpdateModelAnimationBones(model,anim,frame);

    for (int m = 0; m < model.meshCount; m++)
    {
        Mesh mesh = model.meshes[m];
        Vector3 animVertex zeroinit;
        Vector3 animNormal zeroinit;
        int boneId = 0;
        int boneCounter = 0;
        float boneWeight = 0.0;
        bool updated = false; // Flag to check when anim vertex information is updated
        const int vValues = mesh.vertexCount*3;

        // Skip if missing bone data, causes segfault without on some models
        if ((mesh.boneWeights == NULL) || (mesh.boneIds == NULL)) continue;

        for (int vCounter = 0; vCounter < vValues; vCounter += 3)
        {
            mesh.animVertices[vCounter] = 0;
            mesh.animVertices[vCounter + 1] = 0;
            mesh.animVertices[vCounter + 2] = 0;
            if (mesh.animNormals != NULL)
            {
                mesh.animNormals[vCounter] = 0;
                mesh.animNormals[vCounter + 1] = 0;
                mesh.animNormals[vCounter + 2] = 0;
            }

            // Iterates over 4 bones per vertex
            for (int j = 0; j < 4; j++, boneCounter++)
            {
                boneWeight = mesh.boneWeights[boneCounter];
                boneId = mesh.boneIds[boneCounter];

                // Early stop when no transformation will be applied
                if (boneWeight == 0.0f) continue;
                animVertex = CLITERAL(Vector3){ mesh.vertices[vCounter], mesh.vertices[vCounter + 1], mesh.vertices[vCounter + 2] };
                animVertex = Vector3Transform(animVertex,model.meshes[m].boneMatrices[boneId]);
                mesh.animVertices[vCounter] += animVertex.x*boneWeight;
                mesh.animVertices[vCounter+1] += animVertex.y*boneWeight;
                mesh.animVertices[vCounter+2] += animVertex.z*boneWeight;
                updated = true;

                // Normals processing
                // NOTE: We use meshes.baseNormals (default normal) to calculate meshes.normals (animated normals)
                if ((mesh.normals != NULL) && (mesh.animNormals != NULL ))
                {
                    animNormal = CLITERAL(Vector3){ mesh.normals[vCounter], mesh.normals[vCounter + 1], mesh.normals[vCounter + 2] };
                    animNormal = Vector3Transform(animNormal, MatrixTranspose(MatrixInvert(model.meshes[m].boneMatrices[boneId])));
                    mesh.animNormals[vCounter] += animNormal.x*boneWeight;
                    mesh.animNormals[vCounter + 1] += animNormal.y*boneWeight;
                    mesh.animNormals[vCounter + 2] += animNormal.z*boneWeight;
                }
            }
        }

        if (updated)
        {
            BufferData((mesh.vbos[0]), mesh.animVertices, mesh.vertexCount * 3 * sizeof(float)); // Update vertex position
            if (mesh.normals != NULL) BufferData((mesh.vbos[2]), mesh.animNormals, mesh.vertexCount * 3 * sizeof(float)); // Update vertex normals
        }
    }
}
// Load model animations from file
ModelAnimation *LoadModelAnimations(const char *fileName, int *animCount){
    ModelAnimation *animations = NULL;

#if defined(SUPPORT_FILEFORMAT_IQM)
    if (IsFileExtension(fileName, ".iqm")) animations = LoadModelAnimationsIQM(fileName, animCount);
#endif
#if defined(SUPPORT_FILEFORMAT_M3D)
    if (IsFileExtension(fileName, ".m3d")) animations = LoadModelAnimationsM3D(fileName, animCount);
#endif
#if defined(SUPPORT_FILEFORMAT_GLTF)
#endif
    if (IsFileExtension(fileName, ".gltf;.glb")) animations = LoadModelAnimationsGLTF(fileName, animCount);

    return animations;
}

Model LoadModel(const char *fileName){
    
    Model model zeroinit;

    if (IsFileExtension(fileName, ".obj")) model = LoadOBJ(fileName);
    else if (IsFileExtension(fileName, ".glb")) model = LoadGLTF(fileName);
    return model;
}




// 
// 
//  ========== MESH GENERATION ==========
// 
// 




Mesh GenMeshPoly(int sides, float radius)
{
    Mesh mesh zeroinit;

    if (sides < 3) return mesh; // Security check

    int vertexCount = sides*3;

    // Vertices definition
    Vector3 *vertices = (Vector3 *)RL_MALLOC(vertexCount*sizeof(Vector3));

    float d = 0.0f, dStep = 360.0f/sides;
    for (int v = 0; v < vertexCount - 2; v += 3)
    {
        vertices[v] =     CLITERAL(Vector3){ 0.0f, 0.0f, 0.0f };
        vertices[v + 1] = CLITERAL(Vector3){ sinf(DEG2RAD*d)*radius, 0.0f, cosf(DEG2RAD*d)*radius };
        vertices[v + 2] = CLITERAL(Vector3){ sinf(DEG2RAD*(d+dStep))*radius, 0.0f, cosf(DEG2RAD*(d+dStep))*radius };
        d += dStep;
    }

    // Normals definition
    Vector3 *normals = (Vector3 *)RL_MALLOC(vertexCount*sizeof(Vector3));
    for (int n = 0; n < vertexCount; n++) normals[n] = CLITERAL(Vector3){ 0.0f, 1.0f, 0.0f };   // Vector3.up;

    // TexCoords definition
    Vector2 *texcoords = (Vector2 *)RL_MALLOC(vertexCount*sizeof(Vector2));
    for (int n = 0; n < vertexCount; n++) texcoords[n] = CLITERAL(Vector2){ 0.0f, 0.0f };

    mesh.vertexCount = vertexCount;
    mesh.triangleCount = sides;
    mesh.vertices = (float *)RL_MALLOC(mesh.vertexCount*3*sizeof(float));
    mesh.texcoords = (float *)RL_MALLOC(mesh.vertexCount*2*sizeof(float));
    mesh.normals = (float *)RL_MALLOC(mesh.vertexCount*3*sizeof(float));

    // Mesh vertices position array
    for (int i = 0; i < mesh.vertexCount; i++)
    {
        mesh.vertices[3*i] = vertices[i].x;
        mesh.vertices[3*i + 1] = vertices[i].y;
        mesh.vertices[3*i + 2] = vertices[i].z;
    }

    // Mesh texcoords array
    for (int i = 0; i < mesh.vertexCount; i++)
    {
        mesh.texcoords[2*i] = texcoords[i].x;
        mesh.texcoords[2*i + 1] = texcoords[i].y;
    }

    // Mesh normals array
    for (int i = 0; i < mesh.vertexCount; i++)
    {
        mesh.normals[3*i] = normals[i].x;
        mesh.normals[3*i + 1] = normals[i].y;
        mesh.normals[3*i + 2] = normals[i].z;
    }

    RL_FREE(vertices);
    RL_FREE(normals);
    RL_FREE(texcoords);

    // Upload vertex data to GPU (static mesh)
    // NOTE: mesh.vboId array is allocated inside UploadMesh()
    UploadMesh(&mesh, false);

    return mesh;
}

// Generate plane mesh (with subdivisions)
Mesh GenMeshPlane(float width, float length, int resX, int resZ)
{
    Mesh mesh zeroinit;

#define CUSTOM_MESH_GEN_PLANE
#if defined(CUSTOM_MESH_GEN_PLANE)
    resX++;
    resZ++;

    // Vertices definition
    int vertexCount = resX*resZ; // vertices get reused for the faces

    Vector3 *vertices = (Vector3 *)RL_MALLOC(vertexCount*sizeof(Vector3));
    for (int z = 0; z < resZ; z++)
    {
        // [-length/2, length/2]
        float zPos = ((float)z/(resZ - 1) - 0.5f)*length;
        for (int x = 0; x < resX; x++)
        {
            // [-width/2, width/2]
            float xPos = ((float)x/(resX - 1) - 0.5f)*width;
            vertices[x + z*resX] = CLITERAL(Vector3){ xPos, 0.0f, zPos };
        }
    }

    // Normals definition
    Vector3 *normals = (Vector3 *)RL_MALLOC(vertexCount*sizeof(Vector3));
    for (int n = 0; n < vertexCount; n++) normals[n] = CLITERAL(Vector3){ 0.0f, 1.0f, 0.0f };   // Vector3.up;

    // TexCoords definition
    Vector2 *texcoords = (Vector2 *)RL_MALLOC(vertexCount*sizeof(Vector2));
    for (int v = 0; v < resZ; v++)
    {
        for (int u = 0; u < resX; u++)
        {
            texcoords[u + v*resX] = CLITERAL(Vector2){ (float)u/(resX - 1), (float)v/(resZ - 1) };
        }
    }

    // Triangles definition (indices)
    int numFaces = (resX - 1)*(resZ - 1);
    int *triangles = (int *)RL_MALLOC(numFaces*6*sizeof(int));
    int t = 0;
    for (int face = 0; face < numFaces; face++)
    {
        // Retrieve lower left corner from face ind
        int i = face + face/(resX - 1);

        triangles[t++] = i + resX;
        triangles[t++] = i + 1;
        triangles[t++] = i;

        triangles[t++] = i + resX;
        triangles[t++] = i + resX + 1;
        triangles[t++] = i + 1;
    }

    mesh.vertexCount = vertexCount;
    mesh.triangleCount = numFaces*2;
    mesh.vertices = (float *)RL_MALLOC(mesh.vertexCount * 3 * sizeof(float));
    mesh.texcoords = (float *)RL_MALLOC(mesh.vertexCount * 2 * sizeof(float));
    mesh.normals = (float *)RL_MALLOC(mesh.vertexCount * 3 * sizeof(float));
    mesh.indices = (uint32_t*) RL_MALLOC(mesh.triangleCount * 3 * sizeof(uint32_t));

    // Mesh vertices position array
    for (int i = 0; i < mesh.vertexCount; i++)
    {
        mesh.vertices[3*i] = vertices[i].x;
        mesh.vertices[3*i + 1] = vertices[i].y;
        mesh.vertices[3*i + 2] = vertices[i].z;
    }

    // Mesh texcoords array
    for (int i = 0; i < mesh.vertexCount; i++)
    {
        mesh.texcoords[2*i] = texcoords[i].x;
        mesh.texcoords[2*i + 1] = texcoords[i].y;
    }

    // Mesh normals array
    for (int i = 0; i < mesh.vertexCount; i++)
    {
        mesh.normals[3*i] = normals[i].x;
        mesh.normals[3*i + 1] = normals[i].y;
        mesh.normals[3*i + 2] = normals[i].z;
    }

    // Mesh indices array initialization
    for (int i = 0; i < mesh.triangleCount*3; i++) mesh.indices[i] = triangles[i];

    RL_FREE(vertices);
    RL_FREE(normals);
    RL_FREE(texcoords);
    RL_FREE(triangles);

#else       // Use par_shapes library to generate plane mesh

    par_shapes_mesh *plane = par_shapes_create_plane(resX, resZ);   // No normals/texcoords generated!!!
    par_shapes_scale(plane, width, length, 1.0f);
    par_shapes_rotate(plane, -M_PIPI/2.0f,v3_xunit);
    par_shapes_translate(plane, -width/2, 0.0f, length/2);

    mesh.vertices = (float *)RL_MALLOC(plane->ntriangles*3*3*sizeof(float));
    mesh.texcoords = (float *)RL_MALLOC(plane->ntriangles*3*2*sizeof(float));
    mesh.normals = (float *)RL_MALLOC(plane->ntriangles*3*3*sizeof(float));

    mesh.vertexCount = plane->ntriangles*3;
    mesh.triangleCount = plane->ntriangles;

    for (int k = 0; k < mesh.vertexCount; k++)
    {
        mesh.vertices[k*3] = plane->points[plane->triangles[k]*3];
        mesh.vertices[k*3 + 1] = plane->points[plane->triangles[k]*3 + 1];
        mesh.vertices[k*3 + 2] = plane->points[plane->triangles[k]*3 + 2];

        mesh.normals[k*3] = plane->normals[plane->triangles[k]*3];
        mesh.normals[k*3 + 1] = plane->normals[plane->triangles[k]*3 + 1];
        mesh.normals[k*3 + 2] = plane->normals[plane->triangles[k]*3 + 2];

        mesh.texcoords[k*2] = plane->tcoords[plane->triangles[k]*2];
        mesh.texcoords[k*2 + 1] = plane->tcoords[plane->triangles[k]*2 + 1];
    }

    par_shapes_free_mesh(plane);
#endif

    // Upload vertex data to GPU (static mesh)
    UploadMesh(&mesh, false);

    return mesh;
}

// Generated cuboid mesh
Mesh GenMeshCube(float width, float height, float length)
{
    Mesh mesh zeroinit;

#define CUSTOM_MESH_GEN_CUBE
#if defined(CUSTOM_MESH_GEN_CUBE)
    float vertices[] = {
        -width/2, -height/2, length/2,
        width/2, -height/2, length/2,
        width/2, height/2, length/2,
        -width/2, height/2, length/2,
        -width/2, -height/2, -length/2,
        -width/2, height/2, -length/2,
        width/2, height/2, -length/2,
        width/2, -height/2, -length/2,
        -width/2, height/2, -length/2,
        -width/2, height/2, length/2,
        width/2, height/2, length/2,
        width/2, height/2, -length/2,
        -width/2, -height/2, -length/2,
        width/2, -height/2, -length/2,
        width/2, -height/2, length/2,
        -width/2, -height/2, length/2,
        width/2, -height/2, -length/2,
        width/2, height/2, -length/2,
        width/2, height/2, length/2,
        width/2, -height/2, length/2,
        -width/2, -height/2, -length/2,
        -width/2, -height/2, length/2,
        -width/2, height/2, length/2,
        -width/2, height/2, -length/2
    };

    float texcoords[] = {
        0.0f, 0.0f,
        1.0f, 0.0f,
        1.0f, 1.0f,
        0.0f, 1.0f,
        1.0f, 0.0f,
        1.0f, 1.0f,
        0.0f, 1.0f,
        0.0f, 0.0f,
        0.0f, 1.0f,
        0.0f, 0.0f,
        1.0f, 0.0f,
        1.0f, 1.0f,
        1.0f, 1.0f,
        0.0f, 1.0f,
        0.0f, 0.0f,
        1.0f, 0.0f,
        1.0f, 0.0f,
        1.0f, 1.0f,
        0.0f, 1.0f,
        0.0f, 0.0f,
        0.0f, 0.0f,
        1.0f, 0.0f,
        1.0f, 1.0f,
        0.0f, 1.0f
    };

    float normals[] = {
        0.0f, 0.0f, 1.0f,
        0.0f, 0.0f, 1.0f,
        0.0f, 0.0f, 1.0f,
        0.0f, 0.0f, 1.0f,
        0.0f, 0.0f,-1.0f,
        0.0f, 0.0f,-1.0f,
        0.0f, 0.0f,-1.0f,
        0.0f, 0.0f,-1.0f,
        0.0f, 1.0f, 0.0f,
        0.0f, 1.0f, 0.0f,
        0.0f, 1.0f, 0.0f,
        0.0f, 1.0f, 0.0f,
        0.0f,-1.0f, 0.0f,
        0.0f,-1.0f, 0.0f,
        0.0f,-1.0f, 0.0f,
        0.0f,-1.0f, 0.0f,
        1.0f, 0.0f, 0.0f,
        1.0f, 0.0f, 0.0f,
        1.0f, 0.0f, 0.0f,
        1.0f, 0.0f, 0.0f,
        -1.0f, 0.0f, 0.0f,
        -1.0f, 0.0f, 0.0f,
        -1.0f, 0.0f, 0.0f,
        -1.0f, 0.0f, 0.0f
    };

    mesh.vertices = (float *)RL_MALLOC(24*3*sizeof(float));
    memcpy(mesh.vertices, vertices, 24*3*sizeof(float));

    mesh.texcoords = (float *)RL_MALLOC(24*2*sizeof(float));
    memcpy(mesh.texcoords, texcoords, 24*2*sizeof(float));

    mesh.normals = (float *)RL_MALLOC(24*3*sizeof(float));
    memcpy(mesh.normals, normals, 24*3*sizeof(float));

    mesh.indices = (uint32_t *)RL_MALLOC(36*sizeof(uint32_t));

    int k = 0;

    // Indices can be initialized right now
    for (int i = 0; i < 36; i += 6)
    {
        mesh.indices[i] = 4*k;
        mesh.indices[i + 1] = 4*k + 1;
        mesh.indices[i + 2] = 4*k + 2;
        mesh.indices[i + 3] = 4*k;
        mesh.indices[i + 4] = 4*k + 2;
        mesh.indices[i + 5] = 4*k + 3;

        k++;
    }

    mesh.vertexCount = 24;
    mesh.triangleCount = 12;

#else               // Use par_shapes library to generate cube mesh
/*
// Platonic solids:
par_shapes_mesh *par_shapes_create_tetrahedron();       // 4 sides polyhedron (pyramid)
par_shapes_mesh *par_shapes_create_cube();              // 6 sides polyhedron (cube)
par_shapes_mesh *par_shapes_create_octahedron();        // 8 sides polyhedron (diamond)
par_shapes_mesh *par_shapes_create_dodecahedron();      // 12 sides polyhedron
par_shapes_mesh *par_shapes_create_icosahedron();       // 20 sides polyhedron
*/
    // Platonic solid generation: cube (6 sides)
    // NOTE: No normals/texcoords generated by default
    par_shapes_mesh *cube = par_shapes_create_cube();
    cube->tcoords = PAR_MALLOC(float, 2*cube->npoints);
    for (int i = 0; i < 2*cube->npoints; i++) cube->tcoords[i] = 0.0f;
    par_shapes_scale(cube, width, height, length);
    par_shapes_translate(cube, -width/2, 0.0f, -length/2);
    par_shapes_compute_normals(cube);

    mesh.vertices = (float *)RL_MALLOC(cube->ntriangles*3*3*sizeof(float));
    mesh.texcoords = (float *)RL_MALLOC(cube->ntriangles*3*2*sizeof(float));
    mesh.normals = (float *)RL_MALLOC(cube->ntriangles*3*3*sizeof(float));

    mesh.vertexCount = cube->ntriangles*3;
    mesh.triangleCount = cube->ntriangles;

    for (int k = 0; k < mesh.vertexCount; k++)
    {
        mesh.vertices[k*3] = cube->points[cube->triangles[k]*3];
        mesh.vertices[k*3 + 1] = cube->points[cube->triangles[k]*3 + 1];
        mesh.vertices[k*3 + 2] = cube->points[cube->triangles[k]*3 + 2];

        mesh.normals[k*3] = cube->normals[cube->triangles[k]*3];
        mesh.normals[k*3 + 1] = cube->normals[cube->triangles[k]*3 + 1];
        mesh.normals[k*3 + 2] = cube->normals[cube->triangles[k]*3 + 2];

        mesh.texcoords[k*2] = cube->tcoords[cube->triangles[k]*2];
        mesh.texcoords[k*2 + 1] = cube->tcoords[cube->triangles[k]*2 + 1];
    }

    par_shapes_free_mesh(cube);
#endif

    // Upload vertex data to GPU (static mesh)
    UploadMesh(&mesh, false);

    return mesh;
}

// Generate sphere mesh (standard sphere)
Mesh GenMeshSphere(float radius, int rings, int slices)
{
    Mesh mesh zeroinit;

    if ((rings >= 3) && (slices >= 3))
    {
        par_shapes_set_epsilon_degenerate_sphere(0.0);
        par_shapes_mesh *sphere = par_shapes_create_parametric_sphere(slices, rings);
        par_shapes_scale(sphere, radius, radius, radius);
        // NOTE: Soft normals are computed internally

        mesh.vertices = (float *)RL_MALLOC(sphere->ntriangles*3*3*sizeof(float));
        mesh.texcoords = (float *)RL_MALLOC(sphere->ntriangles*3*2*sizeof(float));
        mesh.normals = (float *)RL_MALLOC(sphere->ntriangles*3*3*sizeof(float));

        mesh.vertexCount = sphere->ntriangles*3;
        mesh.triangleCount = sphere->ntriangles;

        for (int k = 0; k < mesh.vertexCount; k++)
        {
            mesh.vertices[k*3] = sphere->points[sphere->triangles[k]*3];
            mesh.vertices[k*3 + 1] = sphere->points[sphere->triangles[k]*3 + 1];
            mesh.vertices[k*3 + 2] = sphere->points[sphere->triangles[k]*3 + 2];

            mesh.normals[k*3] = sphere->normals[sphere->triangles[k]*3];
            mesh.normals[k*3 + 1] = sphere->normals[sphere->triangles[k]*3 + 1];
            mesh.normals[k*3 + 2] = sphere->normals[sphere->triangles[k]*3 + 2];

            mesh.texcoords[k*2] = sphere->tcoords[sphere->triangles[k]*2];
            mesh.texcoords[k*2 + 1] = sphere->tcoords[sphere->triangles[k]*2 + 1];
        }

        par_shapes_free_mesh(sphere);

        // Upload vertex data to GPU (static mesh)
        UploadMesh(&mesh, false);
    }
    else TRACELOG(LOG_WARNING, "MESH: Failed to generate mesh: sphere");

    return mesh;
}

// Generate hemisphere mesh (half sphere, no bottom cap)
Mesh GenMeshHemiSphere(float radius, int rings, int slices)
{
    Mesh mesh zeroinit;

    if ((rings >= 3) && (slices >= 3))
    {
        if (radius < 0.0f) radius = 0.0f;

        par_shapes_mesh *sphere = par_shapes_create_hemisphere(slices, rings);
        par_shapes_scale(sphere, radius, radius, radius);
        // NOTE: Soft normals are computed internally

        mesh.vertices = (float *)RL_MALLOC(sphere->ntriangles*3*3*sizeof(float));
        mesh.texcoords = (float *)RL_MALLOC(sphere->ntriangles*3*2*sizeof(float));
        mesh.normals = (float *)RL_MALLOC(sphere->ntriangles*3*3*sizeof(float));

        mesh.vertexCount = sphere->ntriangles*3;
        mesh.triangleCount = sphere->ntriangles;

        for (int k = 0; k < mesh.vertexCount; k++)
        {
            mesh.vertices[k*3] = sphere->points[sphere->triangles[k]*3];
            mesh.vertices[k*3 + 1] = sphere->points[sphere->triangles[k]*3 + 1];
            mesh.vertices[k*3 + 2] = sphere->points[sphere->triangles[k]*3 + 2];

            mesh.normals[k*3] = sphere->normals[sphere->triangles[k]*3];
            mesh.normals[k*3 + 1] = sphere->normals[sphere->triangles[k]*3 + 1];
            mesh.normals[k*3 + 2] = sphere->normals[sphere->triangles[k]*3 + 2];

            mesh.texcoords[k*2] = sphere->tcoords[sphere->triangles[k]*2];
            mesh.texcoords[k*2 + 1] = sphere->tcoords[sphere->triangles[k]*2 + 1];
        }

        par_shapes_free_mesh(sphere);

        // Upload vertex data to GPU (static mesh)
        UploadMesh(&mesh, false);
    }
    else TRACELOG(LOG_WARNING, "MESH: Failed to generate mesh: hemisphere");

    return mesh;
}

// Generate cylinder mesh
Mesh GenMeshCylinder(float radius, float height, int slices)
{
    Mesh mesh zeroinit;

    if (slices >= 3)
    {
        // Instance a cylinder that sits on the Z=0 plane using the given tessellation
        // levels across the UV domain.  Think of "slices" like a number of pizza
        // slices, and "stacks" like a number of stacked rings
        // Height and radius are both 1.0, but they can easily be changed with par_shapes_scale
        par_shapes_mesh *cylinder = par_shapes_create_cylinder(slices, 8);
        par_shapes_scale(cylinder, radius, radius, height);
        par_shapes_rotate(cylinder, -M_PI/2.0f, v3_xunit);

        // Generate an orientable disk shape (top cap)
        par_shapes_mesh *capTop = par_shapes_create_disk(radius, slices, v3_zero, v3_zunit);
        capTop->tcoords = PAR_MALLOC(float, 2*capTop->npoints);
        for (int i = 0; i < 2*capTop->npoints; i++) capTop->tcoords[i] = 0.0f;
        par_shapes_rotate(capTop, -M_PI/2.0f,v3_xunit);
        par_shapes_rotate(capTop, 90*DEG2RAD,v3_yunit);
        par_shapes_translate(capTop, 0, height, 0);

        // Generate an orientable disk shape (bottom cap)
        par_shapes_mesh *capBottom = par_shapes_create_disk(radius, slices, v3_zero, v3_zunit_negative);
        capBottom->tcoords = PAR_MALLOC(float, 2*capBottom->npoints);
        for (int i = 0; i < 2*capBottom->npoints; i++) capBottom->tcoords[i] = 0.95f;
        par_shapes_rotate(capBottom, M_PI/2.0f,v3_xunit);
        par_shapes_rotate(capBottom, -90*DEG2RAD,v3_yunit);

        par_shapes_merge_and_free(cylinder, capTop);
        par_shapes_merge_and_free(cylinder, capBottom);

        mesh.vertices = (float *)RL_MALLOC(cylinder->ntriangles*3*3*sizeof(float));
        mesh.texcoords = (float *)RL_MALLOC(cylinder->ntriangles*3*2*sizeof(float));
        mesh.normals = (float *)RL_MALLOC(cylinder->ntriangles*3*3*sizeof(float));

        mesh.vertexCount = cylinder->ntriangles*3;
        mesh.triangleCount = cylinder->ntriangles;

        for (int k = 0; k < mesh.vertexCount; k++)
        {
            mesh.vertices[k*3] = cylinder->points[cylinder->triangles[k]*3];
            mesh.vertices[k*3 + 1] = cylinder->points[cylinder->triangles[k]*3 + 1];
            mesh.vertices[k*3 + 2] = cylinder->points[cylinder->triangles[k]*3 + 2];

            mesh.normals[k*3] = cylinder->normals[cylinder->triangles[k]*3];
            mesh.normals[k*3 + 1] = cylinder->normals[cylinder->triangles[k]*3 + 1];
            mesh.normals[k*3 + 2] = cylinder->normals[cylinder->triangles[k]*3 + 2];

            mesh.texcoords[k*2] = cylinder->tcoords[cylinder->triangles[k]*2];
            mesh.texcoords[k*2 + 1] = cylinder->tcoords[cylinder->triangles[k]*2 + 1];
        }

        par_shapes_free_mesh(cylinder);

        // Upload vertex data to GPU (static mesh)
        UploadMesh(&mesh, false);
    }
    else TRACELOG(LOG_WARNING, "MESH: Failed to generate mesh: cylinder");

    return mesh;
}

// Generate cone/pyramid mesh
Mesh GenMeshCone(float radius, float height, int slices)
{
    Mesh mesh zeroinit;

    if (slices >= 3)
    {
        // Instance a cone that sits on the Z=0 plane using the given tessellation
        // levels across the UV domain.  Think of "slices" like a number of pizza
        // slices, and "stacks" like a number of stacked rings
        // Height and radius are both 1.0, but they can easily be changed with par_shapes_scale
        par_shapes_mesh *cone = par_shapes_create_cone(slices, 8);
        par_shapes_scale(cone, radius, radius, height);
        par_shapes_rotate(cone, -M_PI/2.0f,v3_xunit);
        par_shapes_rotate(cone, M_PI/2.0f,v3_yunit);

        // Generate an orientable disk shape (bottom cap)
        par_shapes_mesh *capBottom = par_shapes_create_disk(radius, slices, v3_zero, v3_zunit_negative);
        capBottom->tcoords = PAR_MALLOC(float, 2*capBottom->npoints);
        for (int i = 0; i < 2*capBottom->npoints; i++) capBottom->tcoords[i] = 0.95f;
        par_shapes_rotate(capBottom, M_PI/2.0f,v3_xunit);

        par_shapes_merge_and_free(cone, capBottom);

        mesh.vertices = (float *)RL_MALLOC(cone->ntriangles*3*3*sizeof(float));
        mesh.texcoords = (float *)RL_MALLOC(cone->ntriangles*3*2*sizeof(float));
        mesh.normals = (float *)RL_MALLOC(cone->ntriangles*3*3*sizeof(float));

        mesh.vertexCount = cone->ntriangles*3;
        mesh.triangleCount = cone->ntriangles;

        for (int k = 0; k < mesh.vertexCount; k++)
        {
            mesh.vertices[k*3] = cone->points[cone->triangles[k]*3];
            mesh.vertices[k*3 + 1] = cone->points[cone->triangles[k]*3 + 1];
            mesh.vertices[k*3 + 2] = cone->points[cone->triangles[k]*3 + 2];

            mesh.normals[k*3] = cone->normals[cone->triangles[k]*3];
            mesh.normals[k*3 + 1] = cone->normals[cone->triangles[k]*3 + 1];
            mesh.normals[k*3 + 2] = cone->normals[cone->triangles[k]*3 + 2];

            mesh.texcoords[k*2] = cone->tcoords[cone->triangles[k]*2];
            mesh.texcoords[k*2 + 1] = cone->tcoords[cone->triangles[k]*2 + 1];
        }

        par_shapes_free_mesh(cone);

        // Upload vertex data to GPU (static mesh)
        UploadMesh(&mesh, false);
    }
    else TRACELOG(LOG_WARNING, "MESH: Failed to generate mesh: cone");

    return mesh;
}

// Generate torus mesh
Mesh GenMeshTorus(float radius, float size, int radSeg, int sides)
{
    Mesh mesh zeroinit;

    if ((sides >= 3) && (radSeg >= 3))
    {
        if (radius > 1.0f) radius = 1.0f;
        else if (radius < 0.1f) radius = 0.1f;

        // Create a donut that sits on the Z=0 plane with the specified inner radius
        // The outer radius can be controlled with par_shapes_scale
        par_shapes_mesh *torus = par_shapes_create_torus(radSeg, sides, radius);
        par_shapes_scale(torus, size/2, size/2, size/2);

        mesh.vertices = (float *)RL_MALLOC(torus->ntriangles*3*3*sizeof(float));
        mesh.texcoords = (float *)RL_MALLOC(torus->ntriangles*3*2*sizeof(float));
        mesh.normals = (float *)RL_MALLOC(torus->ntriangles*3*3*sizeof(float));

        mesh.vertexCount = torus->ntriangles*3;
        mesh.triangleCount = torus->ntriangles;

        for (int k = 0; k < mesh.vertexCount; k++)
        {
            mesh.vertices[k*3] = torus->points[torus->triangles[k]*3];
            mesh.vertices[k*3 + 1] = torus->points[torus->triangles[k]*3 + 1];
            mesh.vertices[k*3 + 2] = torus->points[torus->triangles[k]*3 + 2];

            mesh.normals[k*3] = torus->normals[torus->triangles[k]*3];
            mesh.normals[k*3 + 1] = torus->normals[torus->triangles[k]*3 + 1];
            mesh.normals[k*3 + 2] = torus->normals[torus->triangles[k]*3 + 2];

            mesh.texcoords[k*2] = torus->tcoords[torus->triangles[k]*2];
            mesh.texcoords[k*2 + 1] = torus->tcoords[torus->triangles[k]*2 + 1];
        }

        par_shapes_free_mesh(torus);

        // Upload vertex data to GPU (static mesh)
        UploadMesh(&mesh, false);
    }
    else TRACELOG(LOG_WARNING, "MESH: Failed to generate mesh: torus");

    return mesh;
}

// Generate trefoil knot mesh
Mesh GenMeshKnot(float radius, float size, int radSeg, int sides)
{
    Mesh mesh zeroinit;

    if ((sides >= 3) && (radSeg >= 3))
    {
        if (radius > 3.0f) radius = 3.0f;
        else if (radius < 0.5f) radius = 0.5f;

        par_shapes_mesh *knot = par_shapes_create_trefoil_knot(radSeg, sides, radius);
        par_shapes_scale(knot, size, size, size);

        mesh.vertices = (float *)RL_MALLOC(knot->ntriangles*3*3*sizeof(float));
        mesh.texcoords = (float *)RL_MALLOC(knot->ntriangles*3*2*sizeof(float));
        mesh.normals = (float *)RL_MALLOC(knot->ntriangles*3*3*sizeof(float));

        mesh.vertexCount = knot->ntriangles*3;
        mesh.triangleCount = knot->ntriangles;

        for (int k = 0; k < mesh.vertexCount; k++)
        {
            mesh.vertices[k*3] = knot->points[knot->triangles[k]*3];
            mesh.vertices[k*3 + 1] = knot->points[knot->triangles[k]*3 + 1];
            mesh.vertices[k*3 + 2] = knot->points[knot->triangles[k]*3 + 2];

            mesh.normals[k*3] = knot->normals[knot->triangles[k]*3];
            mesh.normals[k*3 + 1] = knot->normals[knot->triangles[k]*3 + 1];
            mesh.normals[k*3 + 2] = knot->normals[knot->triangles[k]*3 + 2];

            mesh.texcoords[k*2] = knot->tcoords[knot->triangles[k]*2];
            mesh.texcoords[k*2 + 1] = knot->tcoords[knot->triangles[k]*2 + 1];
        }

        par_shapes_free_mesh(knot);

        // Upload vertex data to GPU (static mesh)
        UploadMesh(&mesh, false);
    }
    else TRACELOG(LOG_WARNING, "MESH: Failed to generate mesh: knot");

    return mesh;
}

// Generate a mesh from heightmap
// NOTE: Vertex data is uploaded to GPU
Mesh GenMeshHeightmap(Image heightmap, Vector3 size)
{
    #define GRAY_VALUE(c) ((float)(c.r + c.g + c.b)/3.0f)

    Mesh mesh zeroinit;

    int mapX = heightmap.width;
    int mapZ = heightmap.height;

    Color *pixels = LoadImageColors(heightmap);

    // NOTE: One vertex per pixel
    mesh.triangleCount = (mapX - 1)*(mapZ - 1)*2;    // One quad every four pixels

    mesh.vertexCount = mesh.triangleCount*3;

    mesh.vertices = (float *)RL_MALLOC(mesh.vertexCount*3*sizeof(float));
    mesh.normals = (float *)RL_MALLOC(mesh.vertexCount*3*sizeof(float));
    mesh.texcoords = (float *)RL_MALLOC(mesh.vertexCount*2*sizeof(float));
    mesh.colors = NULL;

    int vCounter = 0;       // Used to count vertices float by float
    int tcCounter = 0;      // Used to count texcoords float by float
    int nCounter = 0;       // Used to count normals float by float

    Vector3 scaleFactor = { size.x/(mapX - 1), size.y/255.0f, size.z/(mapZ - 1) };

    Vector3 vA zeroinit;
    Vector3 vB zeroinit;
    Vector3 vC zeroinit;
    Vector3 vN zeroinit;

    for (int z = 0; z < mapZ-1; z++)
    {
        for (int x = 0; x < mapX-1; x++)
        {
            // Fill vertices array with data
            //----------------------------------------------------------

            // one triangle - 3 vertex
            mesh.vertices[vCounter] = (float)x*scaleFactor.x;
            mesh.vertices[vCounter + 1] = GRAY_VALUE(pixels[x + z*mapX])*scaleFactor.y;
            mesh.vertices[vCounter + 2] = (float)z*scaleFactor.z;

            mesh.vertices[vCounter + 3] = (float)x*scaleFactor.x;
            mesh.vertices[vCounter + 4] = GRAY_VALUE(pixels[x + (z + 1)*mapX])*scaleFactor.y;
            mesh.vertices[vCounter + 5] = (float)(z + 1)*scaleFactor.z;

            mesh.vertices[vCounter + 6] = (float)(x + 1)*scaleFactor.x;
            mesh.vertices[vCounter + 7] = GRAY_VALUE(pixels[(x + 1) + z*mapX])*scaleFactor.y;
            mesh.vertices[vCounter + 8] = (float)z*scaleFactor.z;

            // Another triangle - 3 vertex
            mesh.vertices[vCounter + 9] = mesh.vertices[vCounter + 6];
            mesh.vertices[vCounter + 10] = mesh.vertices[vCounter + 7];
            mesh.vertices[vCounter + 11] = mesh.vertices[vCounter + 8];

            mesh.vertices[vCounter + 12] = mesh.vertices[vCounter + 3];
            mesh.vertices[vCounter + 13] = mesh.vertices[vCounter + 4];
            mesh.vertices[vCounter + 14] = mesh.vertices[vCounter + 5];

            mesh.vertices[vCounter + 15] = (float)(x + 1)*scaleFactor.x;
            mesh.vertices[vCounter + 16] = GRAY_VALUE(pixels[(x + 1) + (z + 1)*mapX])*scaleFactor.y;
            mesh.vertices[vCounter + 17] = (float)(z + 1)*scaleFactor.z;
            vCounter += 18;     // 6 vertex, 18 floats

            // Fill texcoords array with data
            //--------------------------------------------------------------
            mesh.texcoords[tcCounter] = (float)x/(mapX - 1);
            mesh.texcoords[tcCounter + 1] = (float)z/(mapZ - 1);

            mesh.texcoords[tcCounter + 2] = (float)x/(mapX - 1);
            mesh.texcoords[tcCounter + 3] = (float)(z + 1)/(mapZ - 1);

            mesh.texcoords[tcCounter + 4] = (float)(x + 1)/(mapX - 1);
            mesh.texcoords[tcCounter + 5] = (float)z/(mapZ - 1);

            mesh.texcoords[tcCounter + 6] = mesh.texcoords[tcCounter + 4];
            mesh.texcoords[tcCounter + 7] = mesh.texcoords[tcCounter + 5];

            mesh.texcoords[tcCounter + 8] = mesh.texcoords[tcCounter + 2];
            mesh.texcoords[tcCounter + 9] = mesh.texcoords[tcCounter + 3];

            mesh.texcoords[tcCounter + 10] = (float)(x + 1)/(mapX - 1);
            mesh.texcoords[tcCounter + 11] = (float)(z + 1)/(mapZ - 1);
            tcCounter += 12;    // 6 texcoords, 12 floats

            // Fill normals array with data
            //--------------------------------------------------------------
            for (int i = 0; i < 18; i += 9)
            {
                vA.x = mesh.vertices[nCounter + i];
                vA.y = mesh.vertices[nCounter + i + 1];
                vA.z = mesh.vertices[nCounter + i + 2];

                vB.x = mesh.vertices[nCounter + i + 3];
                vB.y = mesh.vertices[nCounter + i + 4];
                vB.z = mesh.vertices[nCounter + i + 5];

                vC.x = mesh.vertices[nCounter + i + 6];
                vC.y = mesh.vertices[nCounter + i + 7];
                vC.z = mesh.vertices[nCounter + i + 8];

                vN = Vector3Normalize(Vector3CrossProduct(Vector3Subtract(vB, vA), Vector3Subtract(vC, vA)));

                mesh.normals[nCounter + i] = vN.x;
                mesh.normals[nCounter + i + 1] = vN.y;
                mesh.normals[nCounter + i + 2] = vN.z;

                mesh.normals[nCounter + i + 3] = vN.x;
                mesh.normals[nCounter + i + 4] = vN.y;
                mesh.normals[nCounter + i + 5] = vN.z;

                mesh.normals[nCounter + i + 6] = vN.x;
                mesh.normals[nCounter + i + 7] = vN.y;
                mesh.normals[nCounter + i + 8] = vN.z;
            }

            nCounter += 18;     // 6 vertex, 18 floats
        }
    }

    UnloadImageColors(pixels);  // Unload pixels color data

    // Upload vertex data to GPU (static mesh)
    UploadMesh(&mesh, false);

    return mesh;
}

// Generate a cubes mesh from pixel data
// NOTE: Vertex data is uploaded to GPU
Mesh GenMeshCubicmap(Image cubicmap, Vector3 cubeSize)
{
    #define COLOR_EQUAL(col1, col2) ((col1.r == col2.r)&&(col1.g == col2.g)&&(col1.b == col2.b)&&(col1.a == col2.a))

    Mesh mesh zeroinit;

    Color *pixels = LoadImageColors(cubicmap);

    // NOTE: Max possible number of triangles numCubes*(12 triangles by cube)
    int maxTriangles = cubicmap.width*cubicmap.height*12;

    int vCounter = 0;       // Used to count vertices
    int tcCounter = 0;      // Used to count texcoords
    int nCounter = 0;       // Used to count normals

    float w = cubeSize.x;
    float h = cubeSize.z;
    float h2 = cubeSize.y;

    Vector3 *mapVertices = (Vector3 *)RL_MALLOC(maxTriangles*3*sizeof(Vector3));
    Vector2 *mapTexcoords = (Vector2 *)RL_MALLOC(maxTriangles*3*sizeof(Vector2));
    Vector3 *mapNormals = (Vector3 *)RL_MALLOC(maxTriangles*3*sizeof(Vector3));

    // Define the 6 normals of the cube, we will combine them accordingly later...
    Vector3 n1 = { 1.0f, 0.0f, 0.0f };
    Vector3 n2 = { -1.0f, 0.0f, 0.0f };
    Vector3 n3 = { 0.0f, 1.0f, 0.0f };
    Vector3 n4 = { 0.0f, -1.0f, 0.0f };
    Vector3 n5 = { 0.0f, 0.0f, -1.0f };
    Vector3 n6 = { 0.0f, 0.0f, 1.0f };

    // NOTE: We use texture rectangles to define different textures for top-bottom-front-back-right-left (6)
    typedef struct RectangleF {
        float x;
        float y;
        float width;
        float height;
    } RectangleF;

    RectangleF rightTexUV = { 0.0f, 0.0f, 0.5f, 0.5f };
    RectangleF leftTexUV = { 0.5f, 0.0f, 0.5f, 0.5f };
    RectangleF frontTexUV = { 0.0f, 0.0f, 0.5f, 0.5f };
    RectangleF backTexUV = { 0.5f, 0.0f, 0.5f, 0.5f };
    RectangleF topTexUV = { 0.0f, 0.5f, 0.5f, 0.5f };
    RectangleF bottomTexUV = { 0.5f, 0.5f, 0.5f, 0.5f };

    for (int z = 0; z < cubicmap.height; ++z)
    {
        for (int x = 0; x < cubicmap.width; ++x)
        {
            // Define the 8 vertex of the cube, we will combine them accordingly later...
            Vector3 v1 = { w*(x - 0.5f), h2, h*(z - 0.5f) };
            Vector3 v2 = { w*(x - 0.5f), h2, h*(z + 0.5f) };
            Vector3 v3 = { w*(x + 0.5f), h2, h*(z + 0.5f) };
            Vector3 v4 = { w*(x + 0.5f), h2, h*(z - 0.5f) };
            Vector3 v5 = { w*(x + 0.5f), 0, h*(z - 0.5f) };
            Vector3 v6 = { w*(x - 0.5f), 0, h*(z - 0.5f) };
            Vector3 v7 = { w*(x - 0.5f), 0, h*(z + 0.5f) };
            Vector3 v8 = { w*(x + 0.5f), 0, h*(z + 0.5f) };

            // We check pixel color to be WHITE -> draw full cube
            if (COLOR_EQUAL(pixels[z*cubicmap.width + x], WHITE))
            {
                // Define triangles and checking collateral cubes
                //------------------------------------------------

                // Define top triangles (2 tris, 6 vertex --> v1-v2-v3, v1-v3-v4)
                // WARNING: Not required for a WHITE cubes, created to allow seeing the map from outside
                mapVertices[vCounter] = v1;
                mapVertices[vCounter + 1] = v2;
                mapVertices[vCounter + 2] = v3;
                mapVertices[vCounter + 3] = v1;
                mapVertices[vCounter + 4] = v3;
                mapVertices[vCounter + 5] = v4;
                vCounter += 6;

                mapNormals[nCounter] = n3;
                mapNormals[nCounter + 1] = n3;
                mapNormals[nCounter + 2] = n3;
                mapNormals[nCounter + 3] = n3;
                mapNormals[nCounter + 4] = n3;
                mapNormals[nCounter + 5] = n3;
                nCounter += 6;

                mapTexcoords[tcCounter] = CLITERAL(Vector2){ topTexUV.x, topTexUV.y };
                mapTexcoords[tcCounter + 1] = CLITERAL(Vector2){ topTexUV.x, topTexUV.y + topTexUV.height };
                mapTexcoords[tcCounter + 2] = CLITERAL(Vector2){ topTexUV.x + topTexUV.width, topTexUV.y + topTexUV.height };
                mapTexcoords[tcCounter + 3] = CLITERAL(Vector2){ topTexUV.x, topTexUV.y };
                mapTexcoords[tcCounter + 4] = CLITERAL(Vector2){ topTexUV.x + topTexUV.width, topTexUV.y + topTexUV.height };
                mapTexcoords[tcCounter + 5] = CLITERAL(Vector2){ topTexUV.x + topTexUV.width, topTexUV.y };
                tcCounter += 6;

                // Define bottom triangles (2 tris, 6 vertex --> v6-v8-v7, v6-v5-v8)
                mapVertices[vCounter] = v6;
                mapVertices[vCounter + 1] = v8;
                mapVertices[vCounter + 2] = v7;
                mapVertices[vCounter + 3] = v6;
                mapVertices[vCounter + 4] = v5;
                mapVertices[vCounter + 5] = v8;
                vCounter += 6;

                mapNormals[nCounter] = n4;
                mapNormals[nCounter + 1] = n4;
                mapNormals[nCounter + 2] = n4;
                mapNormals[nCounter + 3] = n4;
                mapNormals[nCounter + 4] = n4;
                mapNormals[nCounter + 5] = n4;
                nCounter += 6;

                mapTexcoords[tcCounter] = CLITERAL(Vector2){ bottomTexUV.x + bottomTexUV.width, bottomTexUV.y };
                mapTexcoords[tcCounter + 1] = CLITERAL(Vector2){ bottomTexUV.x, bottomTexUV.y + bottomTexUV.height };
                mapTexcoords[tcCounter + 2] = CLITERAL(Vector2){ bottomTexUV.x + bottomTexUV.width, bottomTexUV.y + bottomTexUV.height };
                mapTexcoords[tcCounter + 3] = CLITERAL(Vector2){ bottomTexUV.x + bottomTexUV.width, bottomTexUV.y };
                mapTexcoords[tcCounter + 4] = CLITERAL(Vector2){ bottomTexUV.x, bottomTexUV.y };
                mapTexcoords[tcCounter + 5] = CLITERAL(Vector2){ bottomTexUV.x, bottomTexUV.y + bottomTexUV.height };
                tcCounter += 6;

                // Checking cube on bottom of current cube
                if (((z < cubicmap.height - 1) && COLOR_EQUAL(pixels[(z + 1)*cubicmap.width + x], BLACK)) || (z == cubicmap.height - 1))
                {
                    // Define front triangles (2 tris, 6 vertex) --> v2 v7 v3, v3 v7 v8
                    // NOTE: Collateral occluded faces are not generated
                    mapVertices[vCounter] = v2;
                    mapVertices[vCounter + 1] = v7;
                    mapVertices[vCounter + 2] = v3;
                    mapVertices[vCounter + 3] = v3;
                    mapVertices[vCounter + 4] = v7;
                    mapVertices[vCounter + 5] = v8;
                    vCounter += 6;

                    mapNormals[nCounter] = n6;
                    mapNormals[nCounter + 1] = n6;
                    mapNormals[nCounter + 2] = n6;
                    mapNormals[nCounter + 3] = n6;
                    mapNormals[nCounter + 4] = n6;
                    mapNormals[nCounter + 5] = n6;
                    nCounter += 6;

                    mapTexcoords[tcCounter] = CLITERAL(Vector2){ frontTexUV.x, frontTexUV.y };
                    mapTexcoords[tcCounter + 1] = CLITERAL(Vector2){ frontTexUV.x, frontTexUV.y + frontTexUV.height };
                    mapTexcoords[tcCounter + 2] = CLITERAL(Vector2){ frontTexUV.x + frontTexUV.width, frontTexUV.y };
                    mapTexcoords[tcCounter + 3] = CLITERAL(Vector2){ frontTexUV.x + frontTexUV.width, frontTexUV.y };
                    mapTexcoords[tcCounter + 4] = CLITERAL(Vector2){ frontTexUV.x, frontTexUV.y + frontTexUV.height };
                    mapTexcoords[tcCounter + 5] = CLITERAL(Vector2){ frontTexUV.x + frontTexUV.width, frontTexUV.y + frontTexUV.height };
                    tcCounter += 6;
                }

                // Checking cube on top of current cube
                if (((z > 0) && COLOR_EQUAL(pixels[(z - 1)*cubicmap.width + x], BLACK)) || (z == 0))
                {
                    // Define back triangles (2 tris, 6 vertex) --> v1 v5 v6, v1 v4 v5
                    // NOTE: Collateral occluded faces are not generated
                    mapVertices[vCounter] = v1;
                    mapVertices[vCounter + 1] = v5;
                    mapVertices[vCounter + 2] = v6;
                    mapVertices[vCounter + 3] = v1;
                    mapVertices[vCounter + 4] = v4;
                    mapVertices[vCounter + 5] = v5;
                    vCounter += 6;

                    mapNormals[nCounter] = n5;
                    mapNormals[nCounter + 1] = n5;
                    mapNormals[nCounter + 2] = n5;
                    mapNormals[nCounter + 3] = n5;
                    mapNormals[nCounter + 4] = n5;
                    mapNormals[nCounter + 5] = n5;
                    nCounter += 6;

                    mapTexcoords[tcCounter] = CLITERAL(Vector2){ backTexUV.x + backTexUV.width, backTexUV.y };
                    mapTexcoords[tcCounter + 1] = CLITERAL(Vector2){ backTexUV.x, backTexUV.y + backTexUV.height };
                    mapTexcoords[tcCounter + 2] = CLITERAL(Vector2){ backTexUV.x + backTexUV.width, backTexUV.y + backTexUV.height };
                    mapTexcoords[tcCounter + 3] = CLITERAL(Vector2){ backTexUV.x + backTexUV.width, backTexUV.y };
                    mapTexcoords[tcCounter + 4] = CLITERAL(Vector2){ backTexUV.x, backTexUV.y };
                    mapTexcoords[tcCounter + 5] = CLITERAL(Vector2){ backTexUV.x, backTexUV.y + backTexUV.height };
                    tcCounter += 6;
                }

                // Checking cube on right of current cube
                if (((x < cubicmap.width - 1) && COLOR_EQUAL(pixels[z*cubicmap.width + (x + 1)], BLACK)) || (x == cubicmap.width - 1))
                {
                    // Define right triangles (2 tris, 6 vertex) --> v3 v8 v4, v4 v8 v5
                    // NOTE: Collateral occluded faces are not generated
                    mapVertices[vCounter] = v3;
                    mapVertices[vCounter + 1] = v8;
                    mapVertices[vCounter + 2] = v4;
                    mapVertices[vCounter + 3] = v4;
                    mapVertices[vCounter + 4] = v8;
                    mapVertices[vCounter + 5] = v5;
                    vCounter += 6;

                    mapNormals[nCounter] = n1;
                    mapNormals[nCounter + 1] = n1;
                    mapNormals[nCounter + 2] = n1;
                    mapNormals[nCounter + 3] = n1;
                    mapNormals[nCounter + 4] = n1;
                    mapNormals[nCounter + 5] = n1;
                    nCounter += 6;

                    mapTexcoords[tcCounter] = CLITERAL(Vector2){ rightTexUV.x, rightTexUV.y };
                    mapTexcoords[tcCounter + 1] = CLITERAL(Vector2){ rightTexUV.x, rightTexUV.y + rightTexUV.height };
                    mapTexcoords[tcCounter + 2] = CLITERAL(Vector2){ rightTexUV.x + rightTexUV.width, rightTexUV.y };
                    mapTexcoords[tcCounter + 3] = CLITERAL(Vector2){ rightTexUV.x + rightTexUV.width, rightTexUV.y };
                    mapTexcoords[tcCounter + 4] = CLITERAL(Vector2){ rightTexUV.x, rightTexUV.y + rightTexUV.height };
                    mapTexcoords[tcCounter + 5] = CLITERAL(Vector2){ rightTexUV.x + rightTexUV.width, rightTexUV.y + rightTexUV.height };
                    tcCounter += 6;
                }

                // Checking cube on left of current cube
                if (((x > 0) && COLOR_EQUAL(pixels[z*cubicmap.width + (x - 1)], BLACK)) || (x == 0))
                {
                    // Define left triangles (2 tris, 6 vertex) --> v1 v7 v2, v1 v6 v7
                    // NOTE: Collateral occluded faces are not generated
                    mapVertices[vCounter] = v1;
                    mapVertices[vCounter + 1] = v7;
                    mapVertices[vCounter + 2] = v2;
                    mapVertices[vCounter + 3] = v1;
                    mapVertices[vCounter + 4] = v6;
                    mapVertices[vCounter + 5] = v7;
                    vCounter += 6;

                    mapNormals[nCounter] = n2;
                    mapNormals[nCounter + 1] = n2;
                    mapNormals[nCounter + 2] = n2;
                    mapNormals[nCounter + 3] = n2;
                    mapNormals[nCounter + 4] = n2;
                    mapNormals[nCounter + 5] = n2;
                    nCounter += 6;

                    mapTexcoords[tcCounter] = CLITERAL(Vector2){ leftTexUV.x, leftTexUV.y };
                    mapTexcoords[tcCounter + 1] = CLITERAL(Vector2){ leftTexUV.x + leftTexUV.width, leftTexUV.y + leftTexUV.height };
                    mapTexcoords[tcCounter + 2] = CLITERAL(Vector2){ leftTexUV.x + leftTexUV.width, leftTexUV.y };
                    mapTexcoords[tcCounter + 3] = CLITERAL(Vector2){ leftTexUV.x, leftTexUV.y };
                    mapTexcoords[tcCounter + 4] = CLITERAL(Vector2){ leftTexUV.x, leftTexUV.y + leftTexUV.height };
                    mapTexcoords[tcCounter + 5] = CLITERAL(Vector2){ leftTexUV.x + leftTexUV.width, leftTexUV.y + leftTexUV.height };
                    tcCounter += 6;
                }
            }
            // We check pixel color to be BLACK, we will only draw floor and roof
            else if (COLOR_EQUAL(pixels[z*cubicmap.width + x], BLACK))
            {
                // Define top triangles (2 tris, 6 vertex --> v1-v2-v3, v1-v3-v4)
                mapVertices[vCounter] = v1;
                mapVertices[vCounter + 1] = v3;
                mapVertices[vCounter + 2] = v2;
                mapVertices[vCounter + 3] = v1;
                mapVertices[vCounter + 4] = v4;
                mapVertices[vCounter + 5] = v3;
                vCounter += 6;

                mapNormals[nCounter] = n4;
                mapNormals[nCounter + 1] = n4;
                mapNormals[nCounter + 2] = n4;
                mapNormals[nCounter + 3] = n4;
                mapNormals[nCounter + 4] = n4;
                mapNormals[nCounter + 5] = n4;
                nCounter += 6;

                mapTexcoords[tcCounter] = CLITERAL(Vector2){ topTexUV.x, topTexUV.y };
                mapTexcoords[tcCounter + 1] = CLITERAL(Vector2){ topTexUV.x + topTexUV.width, topTexUV.y + topTexUV.height };
                mapTexcoords[tcCounter + 2] = CLITERAL(Vector2){ topTexUV.x, topTexUV.y + topTexUV.height };
                mapTexcoords[tcCounter + 3] = CLITERAL(Vector2){ topTexUV.x, topTexUV.y };
                mapTexcoords[tcCounter + 4] = CLITERAL(Vector2){ topTexUV.x + topTexUV.width, topTexUV.y };
                mapTexcoords[tcCounter + 5] = CLITERAL(Vector2){ topTexUV.x + topTexUV.width, topTexUV.y + topTexUV.height };
                tcCounter += 6;

                // Define bottom triangles (2 tris, 6 vertex --> v6-v8-v7, v6-v5-v8)
                mapVertices[vCounter] = v6;
                mapVertices[vCounter + 1] = v7;
                mapVertices[vCounter + 2] = v8;
                mapVertices[vCounter + 3] = v6;
                mapVertices[vCounter + 4] = v8;
                mapVertices[vCounter + 5] = v5;
                vCounter += 6;

                mapNormals[nCounter] = n3;
                mapNormals[nCounter + 1] = n3;
                mapNormals[nCounter + 2] = n3;
                mapNormals[nCounter + 3] = n3;
                mapNormals[nCounter + 4] = n3;
                mapNormals[nCounter + 5] = n3;
                nCounter += 6;

                mapTexcoords[tcCounter] = CLITERAL(Vector2){ bottomTexUV.x + bottomTexUV.width, bottomTexUV.y };
                mapTexcoords[tcCounter + 1] = CLITERAL(Vector2){ bottomTexUV.x + bottomTexUV.width, bottomTexUV.y + bottomTexUV.height };
                mapTexcoords[tcCounter + 2] = CLITERAL(Vector2){ bottomTexUV.x, bottomTexUV.y + bottomTexUV.height };
                mapTexcoords[tcCounter + 3] = CLITERAL(Vector2){ bottomTexUV.x + bottomTexUV.width, bottomTexUV.y };
                mapTexcoords[tcCounter + 4] = CLITERAL(Vector2){ bottomTexUV.x, bottomTexUV.y + bottomTexUV.height };
                mapTexcoords[tcCounter + 5] = CLITERAL(Vector2){ bottomTexUV.x, bottomTexUV.y };
                tcCounter += 6;
            }
        }
    }

    // Move data from mapVertices temp arrays to vertices float array
    mesh.vertexCount = vCounter;
    mesh.triangleCount = vCounter/3;

    mesh.vertices = (float *)RL_MALLOC(mesh.vertexCount*3*sizeof(float));
    mesh.normals = (float *)RL_MALLOC(mesh.vertexCount*3*sizeof(float));
    mesh.texcoords = (float *)RL_MALLOC(mesh.vertexCount*2*sizeof(float));
    mesh.colors = NULL;

    int fCounter = 0;

    // Move vertices data
    for (int i = 0; i < vCounter; i++)
    {
        mesh.vertices[fCounter] = mapVertices[i].x;
        mesh.vertices[fCounter + 1] = mapVertices[i].y;
        mesh.vertices[fCounter + 2] = mapVertices[i].z;
        fCounter += 3;
    }

    fCounter = 0;

    // Move normals data
    for (int i = 0; i < nCounter; i++)
    {
        mesh.normals[fCounter] = mapNormals[i].x;
        mesh.normals[fCounter + 1] = mapNormals[i].y;
        mesh.normals[fCounter + 2] = mapNormals[i].z;
        fCounter += 3;
    }

    fCounter = 0;

    // Move texcoords data
    for (int i = 0; i < tcCounter; i++)
    {
        mesh.texcoords[fCounter] = mapTexcoords[i].x;
        mesh.texcoords[fCounter + 1] = mapTexcoords[i].y;
        fCounter += 2;
    }

    RL_FREE(mapVertices);
    RL_FREE(mapNormals);
    RL_FREE(mapTexcoords);

    UnloadImageColors(pixels);   // Unload pixels color data

    // Upload vertex data to GPU (static mesh)
    UploadMesh(&mesh, false);

    return mesh;
}