puntorigen
commented
11 hours ago Hi @liw71 ,
Thank you for bringing this up and providing detailed logs! It looks like the issue may be related to the lzma library, which is used for decompressing speaker embedding files. Let’s focus on that first and troubleshoot the problem step by step.
Possible Cause: lzma Library
The _lzma.LZMAError: Corrupt input data error often points to a problem with the lzma library or the integrity of the speaker files being decompressed. Since you’re using Anaconda, it’s possible that the version of lzma bundled with your Python installation is causing the issue.
-
Verify and Reinstall lzma To ensure lzma is properly installed, try reinstalling it using Conda:
conda install -c conda-forge python-lzma -
Test with Fresh Voices The speaker files might have been corrupted during download. To test this, try changing the speaker names in the example script to generate new profiles. For example:
texts = [ {"Speaker1": ["Welcome to the podcast!", "both"]}, {"Speaker2": ["Today, we discuss AI advancements.", "left"]}, ]This will create new voice profiles dynamically and avoid using the existing files. -
Environment Compatibility This package was primarily tested on Python 3.10 in server-like environments (e.g., FastAPI) without using Anaconda. If possible, try creating a clean virtual environment without Anaconda to see if it resolves the issue:
python3.10 -m venv podcast_tts_env source podcast_tts_env/bin/activate pip install podcast_ttsAlternatively, upgrading to Python 3.12 may also help as it has been tested successfully on macOS M1 systems.
Please try the above steps and let me know if the issue persists. If none of these resolve the problem, I can help investigate further or look into potential compatibility issues with the Anaconda setup.
Thank you for your patience, and I appreciate your help in improving podcast_tts!
on my laptop(Macbook m1), I ran into decompress issue when I tried the same demo test,", line 1, in
File "/Users/weili/anaconda3/lib/python3.11/asyncio/runners.py", line 190, in run
return runner.run(main)
^^^^^^^^^^^^^^^^
File "/Users/weili/anaconda3/lib/python3.11/asyncio/runners.py", line 118, in run
return self._loop.run_until_complete(task)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/weili/anaconda3/lib/python3.11/asyncio/base_events.py", line 653, in run_until_complete
return future.result()
^^^^^^^^^^^^^^^
File "", line 3, in main
File "/Users/weili/anaconda3/lib/python3.11/site-packages/podcast_tts/podcast_tts.py", line 314, in generate_wav
wavs = await asyncio.to_thread(
^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/weili/anaconda3/lib/python3.11/asyncio/threads.py", line 25, in to_thread
return await loop.run_in_executor(None, func_call)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/weili/anaconda3/lib/python3.11/concurrent/futures/thread.py", line 58, in run
result = self.fn(*self.args, self.kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/weili/anaconda3/lib/python3.11/site-packages/ChatTTS/core.py", line 221, in infer
return next(res_gen)
^^^^^^^^^^^^^
File "/Users/weili/anaconda3/lib/python3.11/site-packages/ChatTTS/core.py", line 385, in _infer
for result in self._infer_code(
^^^^^^^^^^^^^^^^^
File "/Users/weili/anaconda3/lib/python3.11/site-packages/torch/utils/_contextlib.py", line 116, in decorate_context
return func(*args, *kwargs)
^^^^^^^^^^^^^^^^^^^^^
File "/Users/weili/anaconda3/lib/python3.11/site-packages/ChatTTS/core.py", line 546, in _infer_code
self.speaker.apply(
File "/Users/weili/anaconda3/lib/python3.11/site-packages/torch/utils/_contextlib.py", line 116, in decorate_context
return func(args, kwargs)
^^^^^^^^^^^^^^^^^^^^^
File "/Users/weili/anaconda3/lib/python3.11/site-packages/ChatTTS/model/speaker.py", line 32, in apply
spk_emb_tensor = torch.from_numpy(self._decode(spk_emb))
^^^^^^^^^^^^^^^^^^^^^
File "/Users/weili/anaconda3/lib/python3.11/site-packages/ChatTTS/model/speaker.py", line 148, in _decode
lzma.decompress(
File "/Users/weili/anaconda3/lib/python3.11/lzma.py", line 343, in decompress
res = decomp.decompress(data)
^^^^^^^^^^^^^^^^^^^^^^^
_lzma.LZMAError: Corrupt input data
INFO[0078] #3.3 finished download asset/tokenizer/tokenizer.json INFO[0078] all download tasks finished. Generating audio for chunk 1/1: Hello [uv_break] Welcome to our podcast. text: 0%|▍ | 1/500(max) [00:00, 5.87it/s]We detected that you are passingpast_key_valuesas a tuple of tuples. This is deprecated and will be removed in v4.47. Please convert your cache or use an appropriateCache` class (https://huggingface.co/docs/transformers/kv_cache#legacy-cache-format) text: 2%|████▍ | 11/500(max) [00:00, 15.58it/s] Traceback (most recent call last): File "