Upload folder using huggingface_hub

.gitignore

@@ -14,7 +14,7 @@ dist/
 downloads/
 eggs/
 .eggs/
-# lib/
+lib/
 lib64/
 parts/
 sdist/
@@ -165,4 +165,4 @@ cython_debug/
 #  be found at https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore
 #  and can be added to the global gitignore or merged into this file.  For a more nuclear
 #  option (not recommended) you can uncomment the following to ignore the entire idea folder.
-#.idea/
+#.idea/

README_zh_cn.md

@@ -24,12 +24,12 @@
 
 ## 概览
 
-混元 3D 2.0 是一款先进的大规模 3D 合成系统，用于生成高分辨率的带纹理 3D
-模型。该系统包含两个基础组件：一个大规模形状生成模型 —— 混元 3D-DiT，以及一个大规模纹理合成模型 —— 混元 3D-Paint。
-形状生成模型构建在可扩展的基于流的扩散变换器之上，旨在生成与给定条件图像精确匹配的几何形状，为下游应用奠定坚实基础。
-纹理合成模型得益于强大的几何和扩散先验知识，能够为生成的或手工制作的网格模型生成高分辨率且生动逼真的纹理贴图。
-此外，我们打造了混元 3D 工作室 —— 一个功能多样、易于使用的创作平台，简化了 3D 模型的重建过程。它使专业用户和业余爱好者都能高效地对网格模型进行操作，甚至制作动画。
-我们对模型进行了系统评估，结果表明混元 3D 2.0 在几何细节、条件匹配、纹理质量等方面均优于以往的先进模型，包括开源模型和闭源模型。
+混元 3D 2.0 是一款先进的大规模 3D 资产创作系统，它可以用于生成高分辨率的 3D 白膜以及带纹理的 3D
+模型。该系统包含两个基础组件：一个大规模几何生成模型 — 混元 3D-DiT，以及一个大规模纹理合成模型 — 混元 3D-Paint。
+几何生成模型基于基于流扩散的扩散模型构建，旨在生成与给定条件图像精确匹配的几何模型，为下游应用奠定坚实基础。
+纹理合成模型得益于强大的几何和扩散模型先验知识，能够为AI生成的或手工制作的网格模型生成高分辨率且生动逼真的纹理贴图。
+此外，我们打造了混元 3D 功能矩阵，一个功能多样、易于使用的创作平台，简化了 3D 模型的制作以及修改过程。它使专业用户和业余爱好者都能高效地对3D模型进行操作，甚至制作动画。
+我们对该系统进行了系统评估，结果表明混元 3D 2.0 在几何细节、条件匹配、纹理质量等方面均优于以往的最先进的开源以及闭源模型。
 
 <p align="center">
   <img src="assets/images/system.jpg">
@@ -39,7 +39,7 @@
 
 ### 模型架构
 
-混元 3D 2.0 采用了两阶段生成流程，首先创建一个无纹理的网格模型，然后为该网格模型合成纹理贴图。这种策略有效地将形状生成和纹理生成的难点分离开来，同时也为生成的网格模型或手工制作的网格模型进行纹理处理提供了灵活性。
+混元 3D 2.0 采用了一个两阶段的生成过程，它首先创建一个无纹理的几何模型，然后为该几何模型合成纹理贴图。这种策略有效地将形状生成和纹理生成的难点分离开来，同时也为生成的几何模型或手工制作的几何模型进行纹理处理提供了灵活性。
 
 <p align="left">
   <img src="assets/images/arch.jpg">
@@ -66,10 +66,10 @@
 
 ### 预训练模型
 
-| 模型名称                 | 发布日期       | Huggingface |
-|----------------------|------------|-------------| 
-| Hunyuan3D-DiT-v2-0   | 2025-01-21 | [下载]()      |
-| Hunyuan3D-Paint-v2-0 | 2025-01-21 | [下载]()      |
+| 模型名称                 | 发布日期       | Huggingface                                      |
+|----------------------|------------|--------------------------------------------------| 
+| Hunyuan3D-DiT-v2-0   | 2025-01-21 | [下载](https://huggingface.co/tencent/Hunyuan3D-2) |
+| Hunyuan3D-Paint-v2-0 | 2025-01-21 | [下载](https://huggingface.co/tencent/Hunyuan3D-2) |
 
 ## 🤗快速入门 Hunyuan3D 2.0
 
@@ -85,7 +85,7 @@ pip install -r assets/requirements.txt
 
 ### API 使用方法
 
-我们设计了一个类似于 diffusers 的 API 来使用我们的形状生成模型 —— 混元 3D-DiT 和纹理合成模型 —— 混元 3D-Paint。
+我们设计了一个类似于 diffusers 的 API 来使用我们的几何生成模型 — 混元 3D-DiT 和纹理合成模型 — 混元 3D-Paint。
 你可以通过以下方式使用 混元 3D-DiT：
 
 ```python
@@ -95,7 +95,7 @@ pipeline = Hunyuan3DDiTFlowMatchingPipeline.from_pretrained('tencent/Hunyuan3D-2
 mesh = pipeline(image='assets/demo.png')[0]
 ```
 
-输出的网格是一个trimesh 对象，你可以将其保存为 glb/obj（或其他格式）文件。
+输出的网格是一个 Trimesh 对象，你可以将其保存为 glb/obj（或其他格式）文件。
 对于 混元 3D-Paint，请执行以下操作：
 
 ```python
@@ -110,7 +110,7 @@ pipeline = Hunyuan3DPaintPipeline.from_pretrained('tencent/Hunyuan3D-2')
 mesh = pipeline(mesh, image='assets/demo.png')
 ```
 
-请访问 minimal_demo.py 以了解更多高级用法，例如 文本转 3D 以及 为手工制作的网格生成纹理。
+请访问 [minimal_demo.py](minimal_demo.py) 以了解更多高级用法，例如 文本转 3D 以及 为手工制作的网格生成纹理。
 
 ### Gradio App 使用方法
 
@@ -121,12 +121,13 @@ pip3 install gradio==3.39.0
 python3 gradio_app.py
 ```
 
-如果你不想自己托管，别忘了访问[混元 3D]()进行快速使用。
+如果你不想自己托管，别忘了访问[混元 3D](https://3d.hunyuan.tencent.com)进行快速使用。
 
 ## 📑 开源计划
 
 - [x] 推理代码
 - [x] 模型权重
+- [ ] 技术报告
 - [ ] ComfyUI
 - [ ] TensorRT 量化
 

gradio_app.py

@@ -1,36 +1,15 @@
-# pip install gradio==3.39.0
 import os
-import subprocess
-def install_cuda_toolkit():
-    # CUDA_TOOLKIT_URL = "https://developer.download.nvidia.com/compute/cuda/11.8.0/local_installers/cuda_11.8.0_520.61.05_linux.run"
-    CUDA_TOOLKIT_URL = "https://developer.download.nvidia.com/compute/cuda/12.2.0/local_installers/cuda_12.2.0_535.54.03_linux.run"
-    CUDA_TOOLKIT_FILE = "/tmp/%s" % os.path.basename(CUDA_TOOLKIT_URL)
-    subprocess.call(["wget", "-q", CUDA_TOOLKIT_URL, "-O", CUDA_TOOLKIT_FILE])
-    subprocess.call(["chmod", "+x", CUDA_TOOLKIT_FILE])
-    subprocess.call([CUDA_TOOLKIT_FILE, "--silent", "--toolkit"])
-
-    os.environ["CUDA_HOME"] = "/usr/local/cuda"
-    os.environ["PATH"] = "%s/bin:%s" % (os.environ["CUDA_HOME"], os.environ["PATH"])
-    os.environ["LD_LIBRARY_PATH"] = "%s/lib:%s" % (
-        os.environ["CUDA_HOME"],
-        "" if "LD_LIBRARY_PATH" not in os.environ else os.environ["LD_LIBRARY_PATH"],
-    )
-    # Fix: arch_list[-1] += '+PTX'; IndexError: list index out of range
-    os.environ["TORCH_CUDA_ARCH_LIST"] = "8.0;8.6"
-
-install_cuda_toolkit()
-os.system("cd /home/user/app/hy3dgen/texgen/differentiable_renderer/ && bash compile_mesh_painter.sh")
-os.system("cd /home/user/app/hy3dgen/texgen/custom_rasterizer && pip install .")
-# os.system("cd /home/user/app/hy3dgen/texgen/custom_rasterizer && CUDA_HOME=/usr/local/cuda FORCE_CUDA=1 TORCH_CUDA_ARCH_LIST='8.0;8.6;8.9;9.0' python setup.py install")
-
 import shutil
 import time
 from glob import glob
+from pathlib import Path
+
 import gradio as gr
 import torch
-from gradio_litmodel3d import LitModel3D
+import uvicorn
+from fastapi import FastAPI
+from fastapi.staticfiles import StaticFiles
 
-import spaces
 
 def get_example_img_list():
     print('Loading example img list ...')
@@ -92,7 +71,8 @@ def build_model_viewer_html(save_folder, height=660, width=790, textured=False):
     with open(output_html_path, 'w') as f:
         f.write(template_html.replace('<model-viewer>', obj_html))
 
-    iframe_tag = f'<iframe src="file/{output_html_path}" height="{height}" width="100%" frameborder="0"></iframe>'
+    output_html_path = output_html_path.replace(SAVE_DIR + '/', '')
+    iframe_tag = f'<iframe src="/static/{output_html_path}" height="{height}" width="100%" frameborder="0"></iframe>'
     print(f'Find html {output_html_path}, {os.path.exists(output_html_path)}')
 
     return f"""
@@ -101,7 +81,7 @@ def build_model_viewer_html(save_folder, height=660, width=790, textured=False):
         </div>
     """
 
-@spaces.GPU(duration=60)
+
 def _gen_shape(
     caption,
     image,
@@ -117,10 +97,12 @@ def _gen_shape(
     time_meta = {}
     start_time_0 = time.time()
 
-    image_path = ''
     if image is None:
         start_time = time.time()
-        image = t2i_worker(caption)
+        try:
+            image = t2i_worker(caption)
+        except Exception as e:
+            raise gr.Error(f"Text to 3D is disable. Please enable it by `python gradio_app.py --enable_t23d`.")
         time_meta['text2image'] = time.time() - start_time
 
     image.save(os.path.join(save_folder, 'input.png'))
@@ -158,7 +140,7 @@ def _gen_shape(
     stats['time'] = time_meta
     return mesh, save_folder
 
-@spaces.GPU(duration=80)
+
 def generation_all(
     caption,
     image,
@@ -187,13 +169,11 @@ def generation_all(
     return (
         gr.update(value=path, visible=True),
         gr.update(value=path_textured, visible=True),
-        gr.update(value=path, visible=True),
-        gr.update(value=path_textured, visible=True),
-        # model_viewer_html,
-        # model_viewer_html_textured,
+        model_viewer_html,
+        model_viewer_html_textured,
     )
 
-@spaces.GPU(duration=30)
+
 def shape_generation(
     caption,
     image,
@@ -218,39 +198,28 @@ def shape_generation(
 
     return (
         gr.update(value=path, visible=True),
-        gr.update(value=path, visible=True),
-        # model_viewer_html,
+        model_viewer_html,
     )
 
 
 def build_app():
     title_html = """
-    <div style="font-size: 2em; font-weight: bold; text-align: center; margin-bottom: 20px">
-    
+    <div style="font-size: 2em; font-weight: bold; text-align: center; margin-bottom: 5px">
+
     Hunyuan3D-2: Scaling Diffusion Models for High Resolution Textured 3D Assets Generation
     </div>
     <div align="center">
     Tencent Hunyuan3D Team
     </div>
     <div align="center">
-      <a href="https://github.com/tencent/Hunyuan3D-1">Github Page</a> &ensp; 
+      <a href="https://github.com/tencent/Hunyuan3D-2">Github Page</a> &ensp; 
       <a href="http://3d-models.hunyuan.tencent.com">Homepage</a> &ensp;
-      <a href="https://arxiv.org/pdf/2411.02293">Technical Report</a> &ensp;
+      <a href="#">Technical Report</a> &ensp;
       <a href="https://huggingface.co/Tencent/Hunyuan3D-2"> Models</a> &ensp;
     </div>
     """
-    css = """
-    .json-output {
-        height: 578px;
-    }
-    .json-output .json-holder {
-        height: 538px;
-        overflow-y: scroll;
-    }
-    """
 
-    with gr.Blocks(theme=gr.themes.Base(), css=css, title='Hunyuan-3D-2.0') as demo:
-        # if not gr.__version__.startswith('4'): gr.HTML(title_html)
+    with gr.Blocks(theme=gr.themes.Base(), title='Hunyuan-3D-2.0') as demo:
         gr.HTML(title_html)
 
         with gr.Row():
@@ -261,7 +230,7 @@ def build_app():
                         with gr.Row():
                             check_box_rembg = gr.Checkbox(value=True, label='Remove Background')
 
-                    with gr.Tab('Text Prompt', id='tab_txt_prompt') as tab_tp:
+                    with gr.Tab('Text Prompt', id='tab_txt_prompt', visible=HAS_T2I) as tab_tp:
                         caption = gr.Textbox(label='Text Prompt',
                                              placeholder='HunyuanDiT will be used to generate image.',
                                              info='Example: A 3D model of a cute cat, white background')
@@ -274,7 +243,7 @@ def build_app():
 
                 with gr.Group():
                     btn = gr.Button(value='Generate Shape Only', variant='primary')
-                    btn_all = gr.Button(value='Generate Shape and Texture', variant='primary')
+                    btn_all = gr.Button(value='Generate Shape and Texture', variant='primary', visible=HAS_TEXTUREGEN)
 
                 with gr.Group():
                     file_out = gr.File(label="File", visible=False)
@@ -283,29 +252,9 @@ def build_app():
             with gr.Column(scale=5):
                 with gr.Tabs():
                     with gr.Tab('Generated Mesh') as mesh1:
-                        mesh_output1 = LitModel3D(
-                            label="3D Model1",
-                            exposure=10.0,
-                            height=600,
-                            visible=True,
-                            clear_color=[0.0, 0.0, 0.0, 0.0],
-                            tonemapping="aces",
-                            contrast=1.0,
-                            scale=1.0,
-                        )
-                        # html_output1 = gr.HTML(HTML_OUTPUT_PLACEHOLDER, label='Output')
+                        html_output1 = gr.HTML(HTML_OUTPUT_PLACEHOLDER, label='Output')
                     with gr.Tab('Generated Textured Mesh') as mesh2:
-                        # html_output2 = gr.HTML(HTML_OUTPUT_PLACEHOLDER, label='Output')
-                        mesh_output2 = LitModel3D(
-                            label="3D Model2",
-                            exposure=10.0,
-                            height=600,
-                            visible=True,
-                            clear_color=[0.0, 0.0, 0.0, 0.0],
-                            tonemapping="aces",
-                            contrast=1.0,
-                            scale=1.0,
-                        )
+                        html_output2 = gr.HTML(HTML_OUTPUT_PLACEHOLDER, label='Output')
 
             with gr.Column(scale=2):
                 with gr.Tabs() as gallery:
@@ -314,13 +263,30 @@ def build_app():
                             gr.Examples(examples=example_is, inputs=[image],
                                         label="Image Prompts", examples_per_page=18)
 
-                    with gr.Tab('Text to 3D Gallery', id='tab_txt_gallery') as tab_gt:
+                    with gr.Tab('Text to 3D Gallery', id='tab_txt_gallery', visible=HAS_T2I) as tab_gt:
                         with gr.Row():
                             gr.Examples(examples=example_ts, inputs=[caption],
                                         label="Text Prompts", examples_per_page=18)
 
+        if not HAS_TEXTUREGEN:
+            gr.HTML(""")
+            <div style="margin-top: 20px;">
+                <b>Warning: </b>
+                Texture synthesis is disable due to missing requirements,
+                 please install requirements following README.md to activate it.
+            </div>
+            """)
+        if not args.enable_t23d:
+            gr.HTML("""
+            <div style="margin-top: 20px;">
+                <b>Warning: </b>
+                Text to 3D is disable. To activate it, please run `python gradio_app.py --enable_t23d`.
+            </div>
+            """)
+
         tab_gi.select(fn=lambda: gr.update(selected='tab_img_prompt'), outputs=tabs_prompt)
-        tab_gt.select(fn=lambda: gr.update(selected='tab_txt_prompt'), outputs=tabs_prompt)
+        if HAS_T2I:
+            tab_gt.select(fn=lambda: gr.update(selected='tab_txt_prompt'), outputs=tabs_prompt)
 
         btn.click(
             shape_generation,
@@ -333,8 +299,7 @@ def build_app():
                 octree_resolution,
                 check_box_rembg,
             ],
-            # outputs=[file_out, html_output1]
-            outputs=[file_out, mesh_output1]
+            outputs=[file_out, html_output1]
         ).then(
             lambda: gr.update(visible=True),
             outputs=[file_out],
@@ -351,8 +316,7 @@ def build_app():
                 octree_resolution,
                 check_box_rembg,
             ],
-            # outputs=[file_out, file_out2, html_output1, html_output2]
-            outputs=[file_out, file_out2, mesh_output1, mesh_output2]
+            outputs=[file_out, file_out2, html_output1, html_output2]
         ).then(
             lambda: (gr.update(visible=True), gr.update(visible=True)),
             outputs=[file_out, file_out2],
@@ -366,7 +330,8 @@ if __name__ == '__main__':
 
     parser = argparse.ArgumentParser()
     parser.add_argument('--port', type=int, default=8080)
-    parser.add_argument('--cache-path', type=str, default='./gradio_cache')
+    parser.add_argument('--cache-path', type=str, default='gradio_cache')
+    parser.add_argument('--enable_t23d', action='store_true')
     args = parser.parse_args()
 
     SAVE_DIR = args.cache_path
@@ -386,19 +351,42 @@ if __name__ == '__main__':
     example_is = get_example_img_list()
     example_ts = get_example_txt_list()
 
-    from hy3dgen.text2image import HunyuanDiTPipeline
+    try:
+        from hy3dgen.texgen import Hunyuan3DPaintPipeline
+
+        texgen_worker = Hunyuan3DPaintPipeline.from_pretrained('tencent/Hunyuan3D-2')
+        HAS_TEXTUREGEN = True
+    except Exception as e:
+        print(e)
+        print("Failed to load texture generator.")
+        print('Please try to install requirements by following README.md')
+        HAS_TEXTUREGEN = False
+
+    HAS_T2I = False
+    if args.enable_t23d:
+        from hy3dgen.text2image import HunyuanDiTPipeline
+
+        t2i_worker = HunyuanDiTPipeline('Tencent-Hunyuan/HunyuanDiT-v1.1-Diffusers-Distilled')
+        HAS_T2I = True
+
     from hy3dgen.shapegen import FaceReducer, FloaterRemover, DegenerateFaceRemover, \
         Hunyuan3DDiTFlowMatchingPipeline
-    from hy3dgen.texgen import Hunyuan3DPaintPipeline
     from hy3dgen.rembg import BackgroundRemover
 
     rmbg_worker = BackgroundRemover()
-    t2i_worker = HunyuanDiTPipeline()
     i23d_worker = Hunyuan3DDiTFlowMatchingPipeline.from_pretrained('tencent/Hunyuan3D-2')
-    texgen_worker = Hunyuan3DPaintPipeline.from_pretrained('tencent/Hunyuan3D-2')
     floater_remove_worker = FloaterRemover()
     degenerate_face_remove_worker = DegenerateFaceRemover()
     face_reduce_worker = FaceReducer()
 
+    # https://discuss.huggingface.co/t/how-to-serve-an-html-file/33921/2
+    # create a FastAPI app
+    app = FastAPI()
+    # create a static directory to store the static files
+    static_dir = Path('./gradio_cache')
+    static_dir.mkdir(parents=True, exist_ok=True)
+    app.mount("/static", StaticFiles(directory=static_dir), name="static")
+
     demo = build_app()
-    demo.queue().launch()
+    app = gr.mount_gradio_app(app, demo, path="/")
+    uvicorn.run(app, host="0.0.0.0", port=args.port)

hg_app.py

@@ -0,0 +1,416 @@
+import os
+import spaces
+import subprocess
+def install_cuda_toolkit():
+    # CUDA_TOOLKIT_URL = "https://developer.download.nvidia.com/compute/cuda/11.8.0/local_installers/cuda_11.8.0_520.61.05_linux.run"
+    CUDA_TOOLKIT_URL = "https://developer.download.nvidia.com/compute/cuda/12.2.0/local_installers/cuda_12.2.0_535.54.03_linux.run"
+    CUDA_TOOLKIT_FILE = "/tmp/%s" % os.path.basename(CUDA_TOOLKIT_URL)
+    subprocess.call(["wget", "-q", CUDA_TOOLKIT_URL, "-O", CUDA_TOOLKIT_FILE])
+    subprocess.call(["chmod", "+x", CUDA_TOOLKIT_FILE])
+    subprocess.call([CUDA_TOOLKIT_FILE, "--silent", "--toolkit"])
+
+    os.environ["CUDA_HOME"] = "/usr/local/cuda"
+    os.environ["PATH"] = "%s/bin:%s" % (os.environ["CUDA_HOME"], os.environ["PATH"])
+    os.environ["LD_LIBRARY_PATH"] = "%s/lib:%s" % (
+        os.environ["CUDA_HOME"],
+        "" if "LD_LIBRARY_PATH" not in os.environ else os.environ["LD_LIBRARY_PATH"],
+    )
+    # Fix: arch_list[-1] += '+PTX'; IndexError: list index out of range
+    os.environ["TORCH_CUDA_ARCH_LIST"] = "8.0;8.6"
+
+install_cuda_toolkit()
+os.system("cd /home/user/app/hy3dgen/texgen/differentiable_renderer/ && bash compile_mesh_painter.sh")
+os.system("cd /home/user/app/hy3dgen/texgen/custom_rasterizer && pip install .")
+
+import os
+import shutil
+import time
+from glob import glob
+from pathlib import Path
+
+import gradio as gr
+import torch
+import uvicorn
+from fastapi import FastAPI
+from fastapi.staticfiles import StaticFiles
+
+
+def get_example_img_list():
+    print('Loading example img list ...')
+    return sorted(glob('./assets/example_images/*.png'))
+
+
+def get_example_txt_list():
+    print('Loading example txt list ...')
+    txt_list = list()
+    for line in open('./assets/example_prompts.txt'):
+        txt_list.append(line.strip())
+    return txt_list
+
+
+def gen_save_folder(max_size=60):
+    os.makedirs(SAVE_DIR, exist_ok=True)
+    exists = set(int(_) for _ in os.listdir(SAVE_DIR) if not _.startswith("."))
+    cur_id = min(set(range(max_size)) - exists) if len(exists) < max_size else -1
+    if os.path.exists(f"{SAVE_DIR}/{(cur_id + 1) % max_size}"):
+        shutil.rmtree(f"{SAVE_DIR}/{(cur_id + 1) % max_size}")
+        print(f"remove {SAVE_DIR}/{(cur_id + 1) % max_size} success !!!")
+    save_folder = f"{SAVE_DIR}/{max(0, cur_id)}"
+    os.makedirs(save_folder, exist_ok=True)
+    print(f"mkdir {save_folder} suceess !!!")
+    return save_folder
+
+
+def export_mesh(mesh, save_folder, textured=False):
+    if textured:
+        path = os.path.join(save_folder, f'textured_mesh.glb')
+    else:
+        path = os.path.join(save_folder, f'white_mesh.glb')
+    mesh.export(path, include_normals=textured)
+    return path
+
+
+def build_model_viewer_html(save_folder, height=660, width=790, textured=False):
+    if textured:
+        related_path = f"./textured_mesh.glb"
+        template_name = './assets/modelviewer-textured-template.html'
+        output_html_path = os.path.join(save_folder, f'textured_mesh.html')
+    else:
+        related_path = f"./white_mesh.glb"
+        template_name = './assets/modelviewer-template.html'
+        output_html_path = os.path.join(save_folder, f'white_mesh.html')
+
+    with open(os.path.join(CURRENT_DIR, template_name), 'r') as f:
+        template_html = f.read()
+        obj_html = f"""
+            <div class="column is-mobile is-centered">
+                <model-viewer style="height: {height - 10}px; width: {width}px;" rotation-per-second="10deg" id="modelViewer"
+                    src="{related_path}/" disable-tap 
+                    environment-image="neutral" auto-rotate camera-target="0m 0m 0m" orientation="0deg 0deg 170deg" shadow-intensity=".9"
+                    ar auto-rotate camera-controls>
+                </model-viewer>
+            </div>
+            """
+
+    with open(output_html_path, 'w') as f:
+        f.write(template_html.replace('<model-viewer>', obj_html))
+
+    output_html_path = output_html_path.replace(SAVE_DIR + '/', '')
+    iframe_tag = f'<iframe src="/static/{output_html_path}" height="{height}" width="100%" frameborder="0"></iframe>'
+    print(f'Find html {output_html_path}, {os.path.exists(output_html_path)}')
+
+    return f"""
+        <div style='height: {height}; width: 100%;'>
+        {iframe_tag}
+        </div>
+    """
+
+@spaces.GPU(duration=40)
+def _gen_shape(
+    caption,
+    image,
+    steps=50,
+    guidance_scale=7.5,
+    seed=1234,
+    octree_resolution=256,
+    check_box_rembg=False,
+):
+    if caption: print('prompt is', caption)
+    save_folder = gen_save_folder()
+    stats = {}
+    time_meta = {}
+    start_time_0 = time.time()
+
+    if image is None:
+        start_time = time.time()
+        try:
+            image = t2i_worker(caption)
+        except Exception as e:
+            raise gr.Error(f"Text to 3D is disable. Please enable it by `python gradio_app.py --enable_t23d`.")
+        time_meta['text2image'] = time.time() - start_time
+
+    image.save(os.path.join(save_folder, 'input.png'))
+
+    print(image.mode)
+    if check_box_rembg or image.mode == "RGB":
+        start_time = time.time()
+        image = rmbg_worker(image.convert('RGB'))
+        time_meta['rembg'] = time.time() - start_time
+
+    image.save(os.path.join(save_folder, 'rembg.png'))
+
+    # image to white model
+    start_time = time.time()
+
+    generator = torch.Generator()
+    generator = generator.manual_seed(int(seed))
+    mesh = i23d_worker(
+        image=image,
+        num_inference_steps=steps,
+        guidance_scale=guidance_scale,
+        generator=generator,
+        octree_resolution=octree_resolution
+    )[0]
+
+    mesh = FloaterRemover()(mesh)
+    mesh = DegenerateFaceRemover()(mesh)
+    mesh = FaceReducer()(mesh)
+
+    stats['number_of_faces'] = mesh.faces.shape[0]
+    stats['number_of_vertices'] = mesh.vertices.shape[0]
+
+    time_meta['image_to_textured_3d'] = {'total': time.time() - start_time}
+    time_meta['total'] = time.time() - start_time_0
+    stats['time'] = time_meta
+    return mesh, save_folder
+
+@spaces.GPU(duration=60)
+def generation_all(
+    caption,
+    image,
+    steps=50,
+    guidance_scale=7.5,
+    seed=1234,
+    octree_resolution=256,
+    check_box_rembg=False
+):
+    mesh, save_folder = _gen_shape(
+        caption,
+        image,
+        steps=steps,
+        guidance_scale=guidance_scale,
+        seed=seed,
+        octree_resolution=octree_resolution,
+        check_box_rembg=check_box_rembg
+    )
+    path = export_mesh(mesh, save_folder, textured=False)
+    model_viewer_html = build_model_viewer_html(save_folder, height=596, width=700)
+
+    textured_mesh = texgen_worker(mesh, image)
+    path_textured = export_mesh(textured_mesh, save_folder, textured=True)
+    model_viewer_html_textured = build_model_viewer_html(save_folder, height=596, width=700, textured=True)
+
+    return (
+        gr.update(value=path, visible=True),
+        gr.update(value=path_textured, visible=True),
+        model_viewer_html,
+        model_viewer_html_textured,
+    )
+
+@spaces.GPU(duration=40)
+def shape_generation(
+    caption,
+    image,
+    steps=50,
+    guidance_scale=7.5,
+    seed=1234,
+    octree_resolution=256,
+    check_box_rembg=False,
+):
+    mesh, save_folder = _gen_shape(
+        caption,
+        image,
+        steps=steps,
+        guidance_scale=guidance_scale,
+        seed=seed,
+        octree_resolution=octree_resolution,
+        check_box_rembg=check_box_rembg
+    )
+
+    path = export_mesh(mesh, save_folder, textured=False)
+    model_viewer_html = build_model_viewer_html(save_folder, height=596, width=700)
+
+    return (
+        gr.update(value=path, visible=True),
+        model_viewer_html,
+    )
+
+
+def build_app():
+    title_html = """
+    <div style="font-size: 2em; font-weight: bold; text-align: center; margin-bottom: 5px">
+
+    Hunyuan3D-2: Scaling Diffusion Models for High Resolution Textured 3D Assets Generation
+    </div>
+    <div align="center">
+    Tencent Hunyuan3D Team
+    </div>
+    <div align="center">
+      <a href="https://github.com/tencent/Hunyuan3D-2">Github Page</a> &ensp; 
+      <a href="http://3d-models.hunyuan.tencent.com">Homepage</a> &ensp;
+      <a href="#">Technical Report</a> &ensp;
+      <a href="https://huggingface.co/Tencent/Hunyuan3D-2"> Models</a> &ensp;
+    </div>
+    """
+
+    with gr.Blocks(theme=gr.themes.Base(), title='Hunyuan-3D-2.0') as demo:
+        gr.HTML(title_html)
+
+        with gr.Row():
+            with gr.Column(scale=2):
+                with gr.Tabs() as tabs_prompt:
+                    with gr.Tab('Image Prompt', id='tab_img_prompt') as tab_ip:
+                        image = gr.Image(label='Image', type='pil', image_mode='RGBA', height=290)
+                        with gr.Row():
+                            check_box_rembg = gr.Checkbox(value=True, label='Remove Background')
+
+                    with gr.Tab('Text Prompt', id='tab_txt_prompt', visible=HAS_T2I) as tab_tp:
+                        caption = gr.Textbox(label='Text Prompt',
+                                             placeholder='HunyuanDiT will be used to generate image.',
+                                             info='Example: A 3D model of a cute cat, white background')
+
+                with gr.Accordion('Advanced Options', open=False):
+                    num_steps = gr.Slider(maximum=50, minimum=20, value=30, step=1, label='Inference Steps')
+                    octree_resolution = gr.Dropdown([256, 384, 512], value=256, label='Octree Resolution')
+                    cfg_scale = gr.Number(value=5.5, label='Guidance Scale')
+                    seed = gr.Slider(maximum=1e7, minimum=0, value=1234, label='Seed')
+
+                with gr.Group():
+                    btn = gr.Button(value='Generate Shape Only', variant='primary')
+                    btn_all = gr.Button(value='Generate Shape and Texture', variant='primary', visible=HAS_TEXTUREGEN)
+
+                with gr.Group():
+                    file_out = gr.File(label="File", visible=False)
+                    file_out2 = gr.File(label="File", visible=False)
+
+            with gr.Column(scale=5):
+                with gr.Tabs():
+                    with gr.Tab('Generated Mesh') as mesh1:
+                        html_output1 = gr.HTML(HTML_OUTPUT_PLACEHOLDER, label='Output')
+                    with gr.Tab('Generated Textured Mesh') as mesh2:
+                        html_output2 = gr.HTML(HTML_OUTPUT_PLACEHOLDER, label='Output')
+
+            with gr.Column(scale=2):
+                with gr.Tabs() as gallery:
+                    with gr.Tab('Image to 3D Gallery', id='tab_img_gallery') as tab_gi:
+                        with gr.Row():
+                            gr.Examples(examples=example_is, inputs=[image],
+                                        label="Image Prompts", examples_per_page=18)
+
+                    with gr.Tab('Text to 3D Gallery', id='tab_txt_gallery', visible=HAS_T2I) as tab_gt:
+                        with gr.Row():
+                            gr.Examples(examples=example_ts, inputs=[caption],
+                                        label="Text Prompts", examples_per_page=18)
+
+        if not HAS_TEXTUREGEN:
+            gr.HTML(""")
+            <div style="margin-top: 20px;">
+                <b>Warning: </b>
+                Texture synthesis is disable due to missing requirements,
+                 please install requirements following README.md to activate it.
+            </div>
+            """)
+        if not args.enable_t23d:
+            gr.HTML("""
+            <div style="margin-top: 20px;">
+                <b>Warning: </b>
+                Text to 3D is disable. To activate it, please run `python gradio_app.py --enable_t23d`.
+            </div>
+            """)
+
+        tab_gi.select(fn=lambda: gr.update(selected='tab_img_prompt'), outputs=tabs_prompt)
+        if HAS_T2I:
+            tab_gt.select(fn=lambda: gr.update(selected='tab_txt_prompt'), outputs=tabs_prompt)
+
+        btn.click(
+            shape_generation,
+            inputs=[
+                caption,
+                image,
+                num_steps,
+                cfg_scale,
+                seed,
+                octree_resolution,
+                check_box_rembg,
+            ],
+            outputs=[file_out, html_output1]
+        ).then(
+            lambda: gr.update(visible=True),
+            outputs=[file_out],
+        )
+
+        btn_all.click(
+            generation_all,
+            inputs=[
+                caption,
+                image,
+                num_steps,
+                cfg_scale,
+                seed,
+                octree_resolution,
+                check_box_rembg,
+            ],
+            outputs=[file_out, file_out2, html_output1, html_output2]
+        ).then(
+            lambda: (gr.update(visible=True), gr.update(visible=True)),
+            outputs=[file_out, file_out2],
+        )
+
+    return demo
+
+
+if __name__ == '__main__':
+    import argparse
+
+    parser = argparse.ArgumentParser()
+    parser.add_argument('--port', type=int, default=8080)
+    parser.add_argument('--cache-path', type=str, default='gradio_cache')
+    parser.add_argument('--enable_t23d', default=True)
+    args = parser.parse_args()
+
+    SAVE_DIR = args.cache_path
+    os.makedirs(SAVE_DIR, exist_ok=True)
+
+    CURRENT_DIR = os.path.dirname(os.path.abspath(__file__))
+
+    HTML_OUTPUT_PLACEHOLDER = """
+    <div style='height: 596px; width: 100%; border-radius: 8px; border-color: #e5e7eb; order-style: solid; border-width: 1px;'></div>
+    """
+
+    INPUT_MESH_HTML = """
+    <div style='height: 490px; width: 100%; border-radius: 8px; 
+    border-color: #e5e7eb; order-style: solid; border-width: 1px;'>
+    </div>
+    """
+    example_is = get_example_img_list()
+    example_ts = get_example_txt_list()
+
+    try:
+        from hy3dgen.texgen import Hunyuan3DPaintPipeline
+
+        texgen_worker = Hunyuan3DPaintPipeline.from_pretrained('tencent/Hunyuan3D-2')
+        HAS_TEXTUREGEN = True
+    except Exception as e:
+        print(e)
+        print("Failed to load texture generator.")
+        print('Please try to install requirements by following README.md')
+        HAS_TEXTUREGEN = False
+
+    HAS_T2I = False
+    if args.enable_t23d:
+        from hy3dgen.text2image import HunyuanDiTPipeline
+
+        t2i_worker = HunyuanDiTPipeline('Tencent-Hunyuan/HunyuanDiT-v1.1-Diffusers-Distilled')
+        HAS_T2I = True
+
+    from hy3dgen.shapegen import FaceReducer, FloaterRemover, DegenerateFaceRemover, \
+        Hunyuan3DDiTFlowMatchingPipeline
+    from hy3dgen.rembg import BackgroundRemover
+
+    rmbg_worker = BackgroundRemover()
+    i23d_worker = Hunyuan3DDiTFlowMatchingPipeline.from_pretrained('tencent/Hunyuan3D-2')
+    floater_remove_worker = FloaterRemover()
+    degenerate_face_remove_worker = DegenerateFaceRemover()
+    face_reduce_worker = FaceReducer()
+
+    # https://discuss.huggingface.co/t/how-to-serve-an-html-file/33921/2
+    # create a FastAPI app
+    app = FastAPI()
+    # create a static directory to store the static files
+    static_dir = Path('./gradio_cache')
+    static_dir.mkdir(parents=True, exist_ok=True)
+    app.mount("/static", StaticFiles(directory=static_dir), name="static")
+
+    demo = build_app()
+    app = gr.mount_gradio_app(app, demo, path="/")
+    uvicorn.run(app)

hy3dgen/shapegen/pipelines.py

@@ -209,15 +209,13 @@ class Hunyuan3DDiTPipeline:
         original_model_path = model_path
         if not os.path.exists(model_path):
             # try local path
-            base_dir = os.environ.get('HY3DGEN_MODELS', '/home/user//.cache/hy3dgen')
+            base_dir = os.environ.get('HY3DGEN_MODELS', '~/.cache/hy3dgen')
             model_path = os.path.expanduser(os.path.join(base_dir, model_path, 'hunyuan3d-dit-v2-0'))
             if not os.path.exists(model_path):
                 try:
                     import huggingface_hub
-                    path = huggingface_hub.snapshot_download(
-                        repo_id=original_model_path, 
-                        resume_download=True
-                    )
+                    # download from huggingface
+                    path = huggingface_hub.snapshot_download(repo_id=original_model_path)
                     model_path = os.path.join(path, 'hunyuan3d-dit-v2-0')
                 except ImportError:
                     logger.warning(

hy3dgen/texgen/pipelines.py

@@ -61,7 +61,7 @@ class Hunyuan3DPaintPipeline:
         original_model_path = model_path
         if not os.path.exists(model_path):
             # try local path
-            base_dir = os.environ.get('HY3DGEN_MODELS', '/home/user/.cache/hy3dgen')
+            base_dir = os.environ.get('HY3DGEN_MODELS', '~/.cache/hy3dgen')
             model_path = os.path.expanduser(os.path.join(base_dir, model_path))
 
             delight_model_path = os.path.join(model_path, 'hunyuan3d-delight-v2-0')
@@ -70,12 +70,10 @@ class Hunyuan3DPaintPipeline:
             if not os.path.exists(delight_model_path) or not os.path.exists(multiview_model_path):
                 try:
                     import huggingface_hub
-                    path = huggingface_hub.snapshot_download(
-                        repo_id=original_model_path, 
-                        resume_download=True
-                    )
-                    delight_model_path = os.path.join(path, 'hunyuan3d-delight-v2-0')
-                    multiview_model_path = os.path.join(path, 'hunyuan3d-paint-v2-0')
+                    # download from huggingface
+                    model_path = huggingface_hub.snapshot_download(repo_id=original_model_path)
+                    delight_model_path = os.path.join(model_path, 'hunyuan3d-delight-v2-0')
+                    multiview_model_path = os.path.join(model_path, 'hunyuan3d-paint-v2-0')
                     return cls(Hunyuan3DTexGenConfig(delight_model_path, multiview_model_path))
                 except ImportError:
                     logger.warning(

minimal_demo.py

@@ -26,7 +26,6 @@ import torch
 
 from hy3dgen.rembg import BackgroundRemover
 from hy3dgen.shapegen import Hunyuan3DDiTFlowMatchingPipeline, FaceReducer, FloaterRemover, DegenerateFaceRemover
-from hy3dgen.texgen import Hunyuan3DPaintPipeline
 from hy3dgen.text2image import HunyuanDiTPipeline
 
 
@@ -41,9 +40,14 @@ def image_to_3d(image_path='assets/demo.png'):
     mesh = FaceReducer()(mesh)
     mesh.export('mesh.glb')
 
-    pipeline = Hunyuan3DPaintPipeline.from_pretrained(model_path)
-    mesh = pipeline(mesh, image=image_path)
-    mesh.export('texture.glb')
+    try:
+        from hy3dgen.texgen import Hunyuan3DPaintPipeline
+        pipeline = Hunyuan3DPaintPipeline.from_pretrained(model_path)
+        mesh = pipeline(mesh, image=image_path)
+        mesh.export('texture.glb')
+    except Exception as e:
+        print(e)
+        print('Please try to install requirements by following README.md')
 
 
 def text_to_3d(prompt='a car'):

requirements.txt

@@ -31,3 +31,5 @@ onnxruntime
 pygltflib
 sentencepiece
 gradio
+uvicorn
+fastapi