RuntimeError: Could not load libtorchcodec

[pieraridda934]

Checks

• This template is only for usage issues encountered.
• I have thoroughly reviewed the project documentation but couldn't find information to solve my problem.
• I have searched for existing issues, including closed ones, and couldn't find a solution.
• I am using English to submit this issue to facilitate community communication.

Environment Details

• Windows 11 Pro 25H2
• OS Build 26200.7171
• Python 3.12.10 (tags/v3.12.10:0cc8128, Apr 8 2025, 12:21:36) [MSC v.1943 64 bit (AMD64)] on win32
• ffmpeg version 8.0-full_build-www.gyan.dev

Steps to Reproduce

1. Create a new venv
2. Clone Repo and installed as pip package
3. Run the command "f5-tts_infer-gradio" with nothing provided at cli
4. Attach reference audio and input text to synthesize. Click on Synthesize button and it returns following error

Traceback (most recent call last):
  File "C:\Users\hsn\Python\F5-TTS\venv\Lib\site-packages\gradio\queueing.py", line 759, in process_events
    response = await route_utils.call_process_api(
               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\hsn\Python\F5-TTS\venv\Lib\site-packages\gradio\route_utils.py", line 354, in call_process_api
    output = await app.get_blocks().process_api(
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\hsn\Python\F5-TTS\venv\Lib\site-packages\gradio\blocks.py", line 2191, in process_api
    result = await self.call_function(
             ^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\hsn\Python\F5-TTS\venv\Lib\site-packages\gradio\blocks.py", line 1698, in call_function
    prediction = await anyio.to_thread.run_sync(  # type: ignore
                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\hsn\Python\F5-TTS\venv\Lib\site-packages\anyio\to_thread.py", line 61, in run_sync
    return await get_async_backend().run_sync_in_worker_thread(
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\hsn\Python\F5-TTS\venv\Lib\site-packages\anyio\_backends\_asyncio.py", line 2525, in run_sync_in_worker_thread
    return await future
           ^^^^^^^^^^^^
  File "C:\Users\hsn\Python\F5-TTS\venv\Lib\site-packages\anyio\_backends\_asyncio.py", line 986, in run
    result = context.run(func, *args)
             ^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\hsn\Python\F5-TTS\venv\Lib\site-packages\gradio\utils.py", line 915, in wrapper
    response = f(*args, **kwargs)
               ^^^^^^^^^^^^^^^^^^
  File "C:\Users\hsn\Python\F5-TTS\src\f5_tts\infer\infer_gradio.py", line 290, in basic_tts
    audio_out, spectrogram_path, ref_text_out, used_seed = infer(
                                                           ^^^^^^
  File "C:\Users\hsn\Python\F5-TTS\src\f5_tts\infer\infer_gradio.py", line 160, in infer
    ref_audio, ref_text = preprocess_ref_audio_text(ref_audio_orig, ref_text, show_info=show_info)
                          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\hsn\Python\F5-TTS\src\f5_tts\infer\utils_infer.py", line 361, in preprocess_ref_audio_text
    ref_text = transcribe(ref_audio)
               ^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\hsn\Python\F5-TTS\src\f5_tts\infer\utils_infer.py", line 176, in transcribe
    return asr_pipe(
           ^^^^^^^^^
  File "C:\Users\hsn\Python\F5-TTS\venv\Lib\site-packages\transformers\pipelines\automatic_speech_recognition.py", line 275, in __call__
    return super().__call__(inputs, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\hsn\Python\F5-TTS\venv\Lib\site-packages\transformers\pipelines\base.py", line 1459, in __call__
    return next(
           ^^^^^
  File "C:\Users\hsn\Python\F5-TTS\venv\Lib\site-packages\transformers\pipelines\pt_utils.py", line 126, in __next__
    item = next(self.iterator)
           ^^^^^^^^^^^^^^^^^^^
  File "C:\Users\hsn\Python\F5-TTS\venv\Lib\site-packages\transformers\pipelines\pt_utils.py", line 271, in __next__
    processed = self.infer(next(self.iterator), **self.params)
                           ^^^^^^^^^^^^^^^^^^^
  File "C:\Users\hsn\Python\F5-TTS\venv\Lib\site-packages\torch\utils\data\dataloader.py", line 732, in __next__
    data = self._next_data()
           ^^^^^^^^^^^^^^^^^
  File "C:\Users\hsn\Python\F5-TTS\venv\Lib\site-packages\torch\utils\data\dataloader.py", line 788, in _next_data
    data = self._dataset_fetcher.fetch(index)  # may raise StopIteration
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\hsn\Python\F5-TTS\venv\Lib\site-packages\torch\utils\data\_utils\fetch.py", line 33, in fetch
    data.append(next(self.dataset_iter))
                ^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\hsn\Python\F5-TTS\venv\Lib\site-packages\transformers\pipelines\pt_utils.py", line 188, in __next__
    processed = next(self.subiterator)
                ^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\hsn\Python\F5-TTS\venv\Lib\site-packages\transformers\pipelines\automatic_speech_recognition.py", line 381, in preprocess
    import torchcodec
  File "C:\Users\hsn\Python\F5-TTS\venv\Lib\site-packages\torchcodec\__init__.py", line 12, in <module>
    from . import decoders, encoders, samplers  # noqa
    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\hsn\Python\F5-TTS\venv\Lib\site-packages\torchcodec\decoders\__init__.py", line 7, in <module>
    from .._core import AudioStreamMetadata, VideoStreamMetadata
  File "C:\Users\hsn\Python\F5-TTS\venv\Lib\site-packages\torchcodec\_core\__init__.py", line 8, in <module>
    from ._metadata import (
  File "C:\Users\hsn\Python\F5-TTS\venv\Lib\site-packages\torchcodec\_core\_metadata.py", line 16, in <module>
    from torchcodec._core.ops import (
  File "C:\Users\hsn\Python\F5-TTS\venv\Lib\site-packages\torchcodec\_core\ops.py", line 85, in <module>
    ffmpeg_major_version, core_library_path = load_torchcodec_shared_libraries()
                                              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\hsn\Python\F5-TTS\venv\Lib\site-packages\torchcodec\_core\ops.py", line 70, in load_torchcodec_shared_libraries
    raise RuntimeError(
RuntimeError: Could not load libtorchcodec. Likely causes:
          1. FFmpeg is not properly installed in your environment. We support
             versions 4, 5, 6, 7, and 8.
          2. The PyTorch version (2.9.1+cpu) is not compatible with
             this version of TorchCodec. Refer to the version compatibility
             table:
             https://github.com/pytorch/torchcodec?tab=readme-ov-file#installing-torchcodec.
          3. Another runtime dependency; see exceptions below.
        The following exceptions were raised as we tried to load libtorchcodec:

[start of libtorchcodec loading traceback]
FFmpeg version 8: Could not load this library: C:\Users\hsn\Python\F5-TTS\venv\Lib\site-packages\torchcodec\libtorchcodec_core8.dll
FFmpeg version 7: Could not load this library: C:\Users\hsn\Python\F5-TTS\venv\Lib\site-packages\torchcodec\libtorchcodec_core7.dll
FFmpeg version 6: Could not load this library: C:\Users\hsn\Python\F5-TTS\venv\Lib\site-packages\torchcodec\libtorchcodec_core6.dll
FFmpeg version 5: Could not load this library: C:\Users\hsn\Python\F5-TTS\venv\Lib\site-packages\torchcodec\libtorchcodec_core5.dll
FFmpeg version 4: Could not load this library: C:\Users\hsn\Python\F5-TTS\venv\Lib\site-packages\torchcodec\libtorchcodec_core4.dll
[end of libtorchcodec loading traceback].

✔️ Expected Behavior

output a generated audio

❌ Actual Behavior

Generating error

[SWivid]

Windows 11 Pro 25H2

Try torch<2.9 on Win sys, install cuda compiled torch first before f5tts package/editable:
https://github.com/SWivid/F5-TTS?tab=readme-ov-file#install-pytorch-with-matched-device

[sachin2301]

I downloaded ffmpeg-7.1.1-essentials_build.zip and added the bin directory to the system PATH. After making this change, the issue was resolved.

[amr-hussain]

@pieraridda934 you can uninstall torchcodec and install this version of transformers and it will work for you I guess :

pip uninstall transformers torchcodec -y
pip install transformers==4.38.0

[qiulang]

I think the reason is because torchcodec is listed in https://github.com/SWivid/F5-TTS/blob/main/pyproject.toml#L37

But according to https://github.com/meta-pytorch/torchcodec?tab=readme-ov-file#installing-torchcodec

torchcodec	torch	Python
main / nightly	main / nightly	>=3.10, <=3.14
0.9	2.9	>=3.10, <=3.14
0.8	2.9	>=3.10, <=3.13
0.7	2.8	>=3.9, <=3.13
0.6	2.8	>=3.9, <=3.13
0.5	2.7	>=3.9, <=3.13
0.4	2.7	>=3.9, <=3.13
0.3	2.7	>=3.9, <=3.13
0.2	2.6	>=3.9, <=3.13
0.1	2.5	>=3.9, <=3.12
0.0.3	2.4	>=3.8, <=3.12

As in the current date (2025/12/19) torchcodec 0.9.1 will be installed, which mean torch should be 2.9

But the readme said

# Install pytorch with your CUDA version, e.g.
pip install torch==2.4.0+cu124 torchaudio==2.4.0+cu124 --extra-index-url https://download.pytorch.org/whl/cu124

So torch 2.4 can only use 0.03! That is the case in my environment. I find the F5 source code do not use torchcodec, so I uninstall torchcodec and it works so far.

Does F5 need torchcodec or not. If yes, what version of torchcodec and torch should we use then ?