LatentSync V6版 - 新增批量处理，优化修复，音频驱动视频生成数字人说话视频 本地一键整合包下载

无言***

pzluren 发表于 2025-3-23 11:40
大佬，请问这是什么问题呢？
为什么说人脸检测不到？

人脸模糊，太远/太近都会导致无法检测人脸

AmeCC***

这个项目生成的牙齿像是液化了一样有时会粘连在嘴唇边上，还有牙齿的形状貌似不是固定的，开口闭口间牙齿会不一样？以及内存占用好大，7分半视频能占用200-300gb内存

a44***

debugging consider passing CUDA_LAUNCH_BLOCKING=1\\nCompile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.\\n\',).\n\tWhile copying the parameter named "decoder.blocks.2.cross_attn.out.weight", whose dimensions in the model are torch.Size([384, 384]) and whose dimensions in the checkpoint are torch.Size([384, 384]), an exception occurred : (\'CUDA error: no kernel image is available for execution on the device\\nCUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.\\nFor debugging consider passing CUDA_LAUNCH_BLOCKING=1\\nCompile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.\\n\',).\n\tWhile copying the parameter named "decoder.blocks.2.mlp.0.weight", whose dimensions in the model are torch.Size([1536, 384]) and whose dimensions in the checkpoint are torch.Size([1536, 384]), an exception occurred : (\'CUDA error: no kernel image is available for execution on the device\\nCUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.\\nFor debugging consider passing CUDA_LAUNCH_BLOCKING=1\\nCompile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.\\n\',).\n\tWhile copying the parameter named "decoder.blocks.2.mlp.2.weight", whose dimensions in the model are torch.Size([384, 1536]) and whose dimensions in the checkpoint are torch.Size([384, 1536]), an exception occurred : (\'CUDA error: no kernel image is available for execution on the device\\nCUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.\\nFor debugging consider passing CUDA_LAUNCH_BLOCKING=1\\nCompile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.\\n\',).\n\tWhile copying the parameter named "decoder.blocks.3.attn.query.weight", whose dimensions in the model are torch.Size([384, 384]) and whose dimensions in the checkpoint are torch.Size([384, 384]), an exception occurred : (\'CUDA error: no kernel image is available for execution on the device\\nCUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.\\nFor debugging consider passing CUDA_LAUNCH_BLOCKING=1\\nCompile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.\\n\',).\n\tWhile copying the parameter named "decoder.blocks.3.attn.key.weight", whose dimensions in the model are torch.Size([384, 384]) and whose dimensions in the checkpoint are torch.Size([384, 384]), an exception occurred : (\'CUDA error: no kernel image is available for execution on the device\\nCUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.\\nFor debugging consider passing CUDA_LAUNCH_BLOCKING=1\\nCompile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.\\n\',).\n\tWhile copying the parameter named "decoder.blocks.3.attn.value.weight", whose dimensions in the model are torch.Size([384, 384]) and whose dimensions in the checkpoint are torch.Size([384, 384]), an exception occurred : (\'CUDA error: no kernel image is available for execution on the device\\nCUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.\\nFor debugging consider passing CUDA_LAUNCH_BLOCKING=1\\nCompile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.\\n\',).\n\tWhile copying the parameter named "decoder.blocks.3.attn.out.weight", whose dimensions in the model are torch.Size([384, 384]) and whose dimensions in the checkpoint are torch.Size([384, 384]), an exception occurred : (\'CUDA error: no kernel image is available for execution on the device\\nCUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.\\nFor debugging consider passing CUDA_LAUNCH_BLOCKING=1\\nCompile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.\\n\',).\n\tWhile copying the parameter named "decoder.blocks.3.cross_attn.query.weight", whose dimensions in the model are torch.Size([384, 384]) and whose dimensions in the checkpoint are torch.Size([384, 384]), an exception occurred : (\'CUDA error: no kernel image is available for execution on the device\\nCUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.\\nFor debugging consider passing CUDA_LAUNCH_BLOCKING=1\\nCompile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.\\n\',).\n\tWhile copying the parameter named "decoder.blocks.3.cross_attn.key.weight", whose dimensions in the model are torch.Size([384, 384]) and whose dimensions in the checkpoint are torch.Size([384, 384]), an exception occurred : (\'CUDA error: no kernel image is available for execution on the device\\nCUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.\\nFor debugging consider passing CUDA_LAUNCH_BLOCKING=1\\nCompile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.\\n\',).\n\tWhile copying the parameter named "decoder.blocks.3.cross_attn.value.weight", whose dimensions in the model are torch.Size([384, 384]) and whose dimensions in the checkpoint are torch.Size([384, 384]), an exception occurred : (\'CUDA error: no kernel image is available for execution on the device\\nCUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.\\nFor debugging consider passing CUDA_LAUNCH_BLOCKING=1\\nCompile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.\\n\',).\n\tWhile copying the parameter named "decoder.blocks.3.cross_attn.out.weight", whose dimensions in the model are torch.Size([384, 384]) and whose dimensions in the checkpoint are torch.Size([384, 384]), an exception occurred : (\'CUDA error: no kernel image is available for execution on the device\\nCUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.\\nFor debugging consider passing CUDA_LAUNCH_BLOCKING=1\\nCompile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.\\n\',).\n\tWhile copying the parameter named "decoder.blocks.3.mlp.0.weight", whose dimensions in the model are torch.Size([1536, 384]) and whose dimensions in the checkpoint are torch.Size([1536, 384]), an exception occurred : (\'CUDA error: no kernel image is available for execution on the device\\nCUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.\\nFor debugging consider passing CUDA_LAUNCH_BLOCKING=1\\nCompile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.\\n\',).\n\tWhile copying the parameter named "decoder.blocks.3.mlp.2.weight", whose dimensions in the model are torch.Size([384, 1536]) and whose dimensions in the checkpoint are torch.Size([384, 1536]), an exception occurred : (\'CUDA error: no kernel image is available for execution on the device\\nCUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.\\nFor debugging consider passing CUDA_LAUNCH_BLOCKING=1\\nCompile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.\\n\',).'
大佬，这是什么问题？我搜索说可能是CUDA与PyTorch版本不匹配，但是我用绘图的时候没出现问题。我的显卡是5070ti，cuda版本12.8，python版本3.9，Torch版本2.8.0.dev20250323+cu128。

无言***

a44915 发表于 2025-3-25 21:20
debugging consider passing CUDA_LAUNCH_BLOCKING=1\\nCompile with `TORCH_USE_CUDA_DSA` to enable dev ...

目前我没有50系卡，没法适配50系，全站一键包应该都不支持50系卡

a44***

无言以对 发表于 2025-3-25 21:22
目前我没有50系卡，没法适配50系，全站一键包应该都不支持50系卡

原来是这样

无言***

a44915 发表于 2025-3-26 09:12
原来是这样

有条件你可以自己搭建，torch2.7+CUDA12.8 都是支持50系的

touv***

-----------------------------------------------------
更多实用好玩的AI应用，访问 https://deepface.cc
软件WebUI正在启动中，请稍后...
-----------------------------------------------------
E:\AiTools\LatentSync-V6\deepface\lib\site-packages\_distutils_hack\__init__.py:15: UserWarning: Distutils was imported before Setuptools, but importing Setuptools also replaces the `distutils` module in `sys.modules`. This may lead to undesirable behaviors or errors. To avoid these issues, avoid using distutils directly, ensure that setuptools is installed in the traditional way (e.g. not an editable install), and/or make sure that setuptools is always imported before distutils.
  warnings.warn(
E:\AiTools\LatentSync-V6\deepface\lib\site-packages\_distutils_hack\__init__.py:30: UserWarning: Setuptools is replacing distutils. Support for replacing an already imported distutils is deprecated. In the future, this condition will fail. Register concerns at https://github.com/pypa/setuptoo ... ils-deprecation.yml
  warnings.warn(
信息: 用提供的模式无法找到文件。
* 在本地URL上运行L:  http://127.0.0.1:7860
* 整合包制作：https://deepface.cc

To create a public link, set `share=True` in `launch()`.

启动完后就报了个错。
具体症状就是：所有的解析和生成进度都正常，进度条的发展也正常，最终还能顺利结束，控制台汇报“视频已存入某个路径”。但是却在这个路径下看不到任何文件。temp里有视频和音频的缓存。

touv***

touvidia 发表于 2025-3-29 16:31
-----------------------------------------------------
更多实用好玩的AI应用，访问 https://deepface.c ...

破案了。是我乌龙了。之前的版本产出的视频都保存在一个processed_什么什么路径下。现在的版本改为output了。所以我没细看控制台的输出信息就习惯性地去找那个长长的文件夹名，没找到，就以为文件没生成。。。

事实证明一开始报的错并不影响视频生成。

yb931***

D:\Backup\Pictures\自用\微信截图_20250403145907.pngC:/Users/Administrator.PC-202503171226/AppData/Local/Temp/gradio/f2ba9948f7ea538505ebc8fd9ba41fcb4ff7e20139fd9de3c7a976515d3d576d/4月3日: Invalid argument
Error: Could not open video.
粘贴变换 0 张人脸...
0it [00:00, ?it/s]
处理过程中出错: stack expects a non-empty TensorList
Traceback (most recent call last):
  File "<frozen app>", line 77, in process_video
  File "F:\viode\LatentSync-V6\scripts\inference.py", line 82, in main
    pipeline(
  File "F:\viode\LatentSync-V6\deepface\lib\site-packages\torch\utils\_contextlib.py", line 116, in decorate_context
    return func(*args, **kwargs)
  File "F:\viode\LatentSync-V6\latentsync\pipelines\lipsync_pipeline.py", line 375, in __call__
    video_frames, faces, boxes, affine_matrices = self.loop_video(whisper_chunks, video_frames)
  File "F:\viode\LatentSync-V6\latentsync\pipelines\lipsync_pipeline.py", line 287, in loop_video
    faces, boxes, affine_matrices = self.affine_transform_video(video_frames)
  File "F:\viode\LatentSync-V6\latentsync\pipelines\lipsync_pipeline.py", line 264, in affine_transform_video
    faces = torch.stack(faces)
RuntimeError: stack expects a non-empty TensorList

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "F:\viode\LatentSync-V6\deepface\lib\site-packages\gradio\queueing.py", line 625, in process_events
    response = await route_utils.call_process_api(
  File "F:\viode\LatentSync-V6\deepface\lib\site-packages\gradio\route_utils.py", line 322, in call_process_api
    output = await app.get_blocks().process_api(
  File "F:\viode\LatentSync-V6\deepface\lib\site-packages\gradio\blocks.py", line 2042, in process_api
    result = await self.call_function(
  File "F:\viode\LatentSync-V6\deepface\lib\site-packages\gradio\blocks.py", line 1589, in call_function
    prediction = await anyio.to_thread.run_sync(  # type: ignore
  File "F:\viode\LatentSync-V6\deepface\lib\site-packages\anyio\to_thread.py", line 56, in run_sync
    return await get_async_backend().run_sync_in_worker_thread(
  File "F:\viode\LatentSync-V6\deepface\lib\site-packages\anyio\_backends\_asyncio.py", line 2461, in run_sync_in_worker_thread
    return await future
  File "F:\viode\LatentSync-V6\deepface\lib\site-packages\anyio\_backends\_asyncio.py", line 962, in run
    result = context.run(func, *args)
  File "F:\viode\LatentSync-V6\deepface\lib\site-packages\gradio\utils.py", line 883, in wrapper
    response = f(*args, **kwargs)
  File "<frozen app>", line 96, in process_video
gradio.exceptions.Error: '处理过程中出错: stack expects a non-empty TensorList'
视频输入路径: C:/Users/Administrator.PC-202503171226/AppData/Local/Temp/gradio/6a29810b4eb09f5340ce564a4057a106d5c5bbecb825766cb243f7e3b8c788c0/v4.mp4
音频输入路径: C:/Users/Administrator.PC-202503171226/AppData/Local/Temp/gradio/55d2c3b911b8a3bf107f6a45b2deef89d5469d544b934257e7e818e9997c7dba/4月3日C:/Users/Administrator.PC-202503171226/AppData/Local/Temp/gradio/f2ba9948f7ea538505ebc8fd9ba41fcb4ff7e20139fd9de3c7a976515d3d576d/4月3日: Invalid argument
Error: Could not open video.
粘贴变换 0 张人脸...
0it [00:00, ?it/s]
处理过程中出错: stack expects a non-empty TensorList
Traceback (most recent call last):
  File "<frozen app>", line 77, in process_video
  File "F:\viode\LatentSync-V6\scripts\inference.py", line 82, in main
    pipeline(
  File "F:\viode\LatentSync-V6\deepface\lib\site-packages\torch\utils\_contextlib.py", line 116, in decorate_context
    return func(*args, **kwargs)
  File "F:\viode\LatentSync-V6\latentsync\pipelines\lipsync_pipeline.py", line 375, in __call__
    video_frames, faces, boxes, affine_matrices = self.loop_video(whisper_chunks, video_frames)
  File "F:\viode\LatentSync-V6\latentsync\pipelines\lipsync_pipeline.py", line 287, in loop_video
    faces, boxes, affine_matrices = self.affine_transform_video(video_frames)
  File "F:\viode\LatentSync-V6\latentsync\pipelines\lipsync_pipeline.py", line 264, in affine_transform_video
    faces = torch.stack(faces)
RuntimeError: stack expects a non-empty TensorList

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "F:\viode\LatentSync-V6\deepface\lib\site-packages\gradio\queueing.py", line 625, in process_events
    response = await route_utils.call_process_api(
  File "F:\viode\LatentSync-V6\deepface\lib\site-packages\gradio\route_utils.py", line 322, in call_process_api
    output = await app.get_blocks().process_api(
  File "F:\viode\LatentSync-V6\deepface\lib\site-packages\gradio\blocks.py", line 2042, in process_api
    result = await self.call_function(
  File "F:\viode\LatentSync-V6\deepface\lib\site-packages\gradio\blocks.py", line 1589, in call_function
    prediction = await anyio.to_thread.run_sync(  # type: ignore
  File "F:\viode\LatentSync-V6\deepface\lib\site-packages\anyio\to_thread.py", line 56, in run_sync
    return await get_async_backend().run_sync_in_worker_thread(
  File "F:\viode\LatentSync-V6\deepface\lib\site-packages\anyio\_backends\_asyncio.py", line 2461, in run_sync_in_worker_thread
    return await future
  File "F:\viode\LatentSync-V6\deepface\lib\site-packages\anyio\_backends\_asyncio.py", line 962, in run
    result = context.run(func, *args)
  File "F:\viode\LatentSync-V6\deepface\lib\site-packages\gradio\utils.py", line 883, in wrapper
    response = f(*args, **kwargs)
  File "<frozen app>", line 96, in process_video
gradio.exceptions.Error: '处理过程中出错: stack expects a non-empty TensorList'
视频输入路径: C:/Users/Administrator.PC-202503171226/AppData/Local/Temp/gradio/6a29810b4eb09f5340ce564a4057a106d5c5bbecb825766cb243f7e3b8c788c0/v4.mp4
音频输入路径: C:/Users/Administrator.PC-202503171226/AppData/Local/Temp/gradio/55d2c3b911b8a3bf107f6a45b2deef89d5469d544b934257e7e818e9997c7dba/4月3日
这个怎么处理

yb931***

ConnectionResetError: [WinError 10054] 远程主机强迫关闭了一个现有的连接。
视频输入路径: C:/Users/Administrator.PC-202503171226/AppData/Local/Temp/gradio/f2ba9948f7ea538505ebc8fd9ba41fcb4ff7e20139fd9de3c7a976515d3d576d/4月3日 11.mp4
音频输入路径: C:/Users/Administrator.PC-202503171226/AppData/Local/Temp/gradio/55d2c3b911b8a3bf107f6a45b2deef89d5469d544b934257e7e818e9997c7dba/4月3日 1.WAV
从以下路径加载模型: F:/viode/LatentSync-V6/checkpoints/latentsync_unet.pt
初始化种子: 1247
视频帧率为 25 FPS, 音频idx帧率 50FPS
C:/Users/Administrator.PC-202503171226/AppData/Local/Temp/gradio/f2ba9948f7ea538505ebc8fd9ba41fcb4ff7e20139fd9de3c7a976515d3d576d/4月3日: Invalid argument
Error: Could not open video.
粘贴变换 0 张人脸...
0it [00:00, ?it/s]
处理过程中出错: stack expects a non-empty TensorList
Traceback (most recent call last):
  File "<frozen app>", line 77, in process_video
  File "F:\viode\LatentSync-V6\scripts\inference.py", line 82, in main
    pipeline(
  File "F:\viode\LatentSync-V6\deepface\lib\site-packages\torch\utils\_contextlib.py", line 116, in decorate_context
    return func(*args, **kwargs)
  File "F:\viode\LatentSync-V6\latentsync\pipelines\lipsync_pipeline.py", line 375, in __call__
    video_frames, faces, boxes, affine_matrices = self.loop_video(whisper_chunks, video_frames)
  File "F:\viode\LatentSync-V6\latentsync\pipelines\lipsync_pipeline.py", line 287, in loop_video
    faces, boxes, affine_matrices = self.affine_transform_video(video_frames)
  File "F:\viode\LatentSync-V6\latentsync\pipelines\lipsync_pipeline.py", line 264, in affine_transform_video
    faces = torch.stack(faces)
RuntimeError: stack expects a non-empty TensorList

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "F:\viode\LatentSync-V6\deepface\lib\site-packages\gradio\queueing.py", line 625, in process_events
    response = await route_utils.call_process_api(
  File "F:\viode\LatentSync-V6\deepface\lib\site-packages\gradio\route_utils.py", line 322, in call_process_api
    output = await app.get_blocks().process_api(
  File "F:\viode\LatentSync-V6\deepface\lib\site-packages\gradio\blocks.py", line 2042, in process_api
    result = await self.call_function(
  File "F:\viode\LatentSync-V6\deepface\lib\site-packages\gradio\blocks.py", line 1589, in call_function
    prediction = await anyio.to_thread.run_sync(  # type: ignore
  File "F:\viode\LatentSync-V6\deepface\lib\site-packages\anyio\to_thread.py", line 56, in run_sync
    return await get_async_backend().run_sync_in_worker_thread(
  File "F:\viode\LatentSync-V6\deepface\lib\site-packages\anyio\_backends\_asyncio.py", line 2461, in run_sync_in_worker_thread
    return await future
  File "F:\viode\LatentSync-V6\deepface\lib\site-packages\anyio\_backends\_asyncio.py", line 962, in run
    result = context.run(func, *args)
  File "F:\viode\LatentSync-V6\deepface\lib\site-packages\gradio\utils.py", line 883, in wrapper
    response = f(*args, **kwargs)
  File "<frozen app>", line 96, in process_video
gradio.exceptions.Error: '处理过程中出错: stack expects a non-empty TensorList'
大神，求解答