一、前言

网上的教程均为搭配opengl使用，如果单纯想读取模型数据，资料就比较少了。先放出相关链接：

1、gltf规范文档：glTF™ 2.0 Specification (khronos.org)

2、gltf在线模型查看器 ：glTF Viewer (donmccurdy.com)

3、gltf>tinygltf：GitHub – syoyo/gltf>tinygltf: Header only C++11 tiny glTF 2.0 library

4、vulkan使用gltf>tinygltf官方示例：GitHub – SaschaWillems/Vulkan-glTF-PBR: Physical based rendering with Vulkan using glTF 2.0 models

 二、使用

1、初始化

首先引入头文件：

#define TINYGLTF_IMPLEMENTATION
#include <tiny_gltf.h>

再加载文件，注意glb格式和gltf格式调用的函数不同：

gltf>tinygltf::Model model;
gltf>tinygltf::TinyGLTF loader;
std::string err;
std::string warn;bool load_ok;
if(is_glb)load_ok = loader.LoadBinaryFromFile(&model, &err, &warn, std::string{ path_name });
elseload_ok = loader.LoadASCIIFromFile(&model, &err, &warn, std::string{ path_name });if (!warn.empty()) log_warn("{}", warn);if (!err.empty())log_err("{}", err);if (!load_ok)
{log_err("解析gltf格式失败：{}", err);return {};
}

二、读取材质

其中 mo_material是我们自定义的结构，用于接收读取到的数据。由于之前的光照模型是 Blinn-Phong，而现在更通用的是pbr，以下代码可能不适用你的需求。

之所以我将纹理导出、因为我要进行二次处理，生成自定义的模型格式。你可以不保存到文件，而是直接使用。

通过查看 gltf>tinygltf::PbrMetallicRoughness类型的定义，可以知道提供哪些纹理和材质数据。

//------------------获取材质-------------------
mo_material.resize(model.materials.size());
for (size_t i = 0; i < model.materials.size(); ++i)
{gltf>tinygltf::Material& m = model.materials[i];log_info("材质：{}，{}", i, m.name);gltf>tinygltf::PbrMetallicRoughness& pbr = m.pbrMetallicRoughness;ModelObjectMaterial& material = mo_material[i];material._materialName = m.name;material._material._diffuseAlbedo = { (float)pbr.baseColorFactor[0],(float)pbr.baseColorFactor[1],(float)pbr.baseColorFactor[2], 1.0f };material._material._fresnelR0 = { (float)pbr.metallicFactor, (float)pbr.metallicFactor, (float)pbr.metallicFactor };material._material._roughness = (float)pbr.roughnessFactor;// 纹理导出auto fn_texture = [&](int index, std::string_view tag) {if (index == -1)return std::string{};gltf>tinygltf::Image& img = model.images[index];if (!img.uri.empty())return img.uri;std::string path_name = fmt::format("{}{}{}#{}.png",File::PathTemp(), res_name, tag, i);stbi_write_png(CodeCvt::Utf8ToMultiByte(path_name).c_str(),img.width, img.height,4, img.image.data(), 0);return path_name;};material._pathTexDiffuse = fn_texture(pbr.baseColorTexture.index, "");material._pathTexNormal = fn_texture(m.normalTexture.index, "_normal");
}

我的做法是按材质合并网格（不知道会不会影响骨骼动画，后面再说），你也可以简单处理，按 gltf>tinygltf::Primitive为单位为绘制。prim里不直接存数据，而是通过 Accessor访问，然后通过 Accessor取得 BufferView，然后通过 BufferView取得真正的顶点数据。

注意数据首地址为 buf.data.data() + buf_view.byteOffset + accessor.byteOffset，单个顶点偏移为 int byte_stride = accessor.ByteStride(buf_view)。

注意我这里没有过多的 assert的检查，这里的代码还需要尝试更多模型，才能具有可靠性。

		for (gltf>tinygltf::Mesh& mesh : model.meshes){for (gltf>tinygltf::Primitive& prim : mesh.primitives){PrimData& prim_data = map_material[prim.material];size_t vertex_offset = prim_data._allVertex.size();// 索引{gltf>tinygltf::Accessor accessor = model.accessors[prim.indices];gltf>tinygltf::BufferView buf_view = model.bufferViews[accessor.bufferView];gltf>tinygltf::Buffer& buf = model.buffers[buf_view.buffer];int byte_stride = accessor.ByteStride(buf_view);unsigned char* p = buf.data.data() + buf_view.byteOffset + accessor.byteOffset;assert(byte_stride == 4);for (size_t i = 0; i < accessor.count; ++i){uint32_t index = *(uint32_t*)p; prim_data._allIndex.push_back(vertex_offset + index);p += byte_stride;}}// 位置auto iter_pos = prim.attributes.find("POSITION");if(iter_pos != prim.attributes.end()){gltf>tinygltf::Accessor accessor = model.accessors[iter_pos->second];gltf>tinygltf::BufferView buf_view = model.bufferViews[accessor.bufferView];gltf>tinygltf::Buffer& buf = model.buffers[buf_view.buffer];prim_data._allVertex.resize(vertex_offset + accessor.count);int byte_stride = accessor.ByteStride(buf_view);unsigned char* p = buf.data.data() + buf_view.byteOffset + accessor.byteOffset;for (size_t i = 0; i < accessor.count; ++i){Vertex& v = prim_data._allVertex[vertex_offset + i];v._pos = *(Position3*)p;p += byte_stride;}}// uvauto iter_uv = prim.attributes.find("TEXCOORD_0");if (iter_uv != prim.attributes.end()){gltf>tinygltf::Accessor accessor = model.accessors[iter_uv->second];gltf>tinygltf::BufferView buf_view = model.bufferViews[accessor.bufferView];gltf>tinygltf::Buffer& buf = model.buffers[buf_view.buffer];int byte_stride = accessor.ByteStride(buf_view);unsigned char* p = buf.data.data() + buf_view.byteOffset + accessor.byteOffset;for (size_t i = 0; i < accessor.count; ++i){Vertex& v = prim_data._allVertex[vertex_offset + i];v._uv = *(Position2*)p;p += byte_stride;}}// normalauto iter_normal = prim.attributes.find("NORMAL");if (iter_normal != prim.attributes.end()){gltf>tinygltf::Accessor accessor = model.accessors[iter_normal->second];gltf>tinygltf::BufferView buf_view = model.bufferViews[accessor.bufferView];gltf>tinygltf::Buffer& buf = model.buffers[buf_view.buffer];int byte_stride = accessor.ByteStride(buf_view);unsigned char* p = buf.data.data() + buf_view.byteOffset + accessor.byteOffset;for (size_t i = 0; i < accessor.count; ++i){Vertex& v = prim_data._allVertex[vertex_offset + i];v._normal = *(Position3*)p;p += byte_stride;}}// TANGENTauto iter_tangent = prim.attributes.find("TANGENT");if (iter_tangent != prim.attributes.end()){gltf>tinygltf::Accessor accessor = model.accessors[iter_tangent->second];gltf>tinygltf::BufferView buf_view = model.bufferViews[accessor.bufferView];gltf>tinygltf::Buffer& buf = model.buffers[buf_view.buffer];int byte_stride = accessor.ByteStride(buf_view);unsigned char* p = buf.data.data() + buf_view.byteOffset + accessor.byteOffset;for (size_t i = 0; i < accessor.count; ++i){Vertex& v = prim_data._allVertex[vertex_offset + i];v._tangent = *(Position3*)p;p += byte_stride;}}}}

三、截图

这里没有处理骨骼动画和光照，不过对于学习 gltf>tinygltf的基本用法，应该是足够了！

项目地址：dl: C++ drawing library (gitee.com)