Wav2Lip Collab not working anymore -| TypeError: mel() takes 0 positional arguments

[acetra19]

Hey folks, unfortunately for seemingly no obvious reason at all,

I can not get the wav2lip-collab to work anymore. I never had any problems until a couple of days ago.

Now I always get the following error on the last step (rendering):

Using cuda for inference. Reading video frames... Number of frames available for inference: 353 Traceback (most recent call last): File "/content/Wav2Lip/inference.py", line 280, in main() File "/content/Wav2Lip/inference.py", line 225, in main mel = audio.melspectrogram(wav) File "/content/Wav2Lip/audio.py", line 47, in melspectrogram S = _amp_to_db(_linear_to_mel(np.abs(D))) - hp.ref_level_db File "/content/Wav2Lip/audio.py", line 95, in _linear_to_mel _mel_basis = _build_mel_basis() File "/content/Wav2Lip/audio.py", line 100, in _build_mel_basis return librosa.filters.mel(hp.sample_rate, hp.n_fft, n_mels=hp.num_mels, TypeError: mel() takes 0 positional arguments but 2 positional arguments (and 3 keyword-only arguments) were given

any idea how to troubleshoot this problem? Any help would be appreciated.

Best regards!

[bobwatcherx]

i also get error messsage

even though the duration of the video with sound is the same as 7 seconds

Using cuda for inference.
Reading video frames...
Number of frames available for inference: 1693
Traceback (most recent call last):
  File "/content/Wav2Lip/inference.py", line 280, in <module>
    main()
  File "/content/Wav2Lip/inference.py", line 225, in main
    mel = audio.melspectrogram(wav)
  File "/content/Wav2Lip/audio.py", line 47, in melspectrogram
    S = _amp_to_db(_linear_to_mel(np.abs(D))) - hp.ref_level_db
  File "/content/Wav2Lip/audio.py", line 95, in _linear_to_mel
    _mel_basis = _build_mel_basis()
  File "/content/Wav2Lip/audio.py", line 100, in _build_mel_basis
    return librosa.filters.mel(hp.sample_rate, hp.n_fft, n_mels=hp.num_mels,
TypeError: mel() takes 0 positional arguments but 2 positional arguments (and 3 keyword-only arguments) were given

[bobwatcherx]

Hey folks, unfortunately for seemingly no obvious reason at all,

I can not get the wav2lip-collab to work anymore. I never had any problems until a couple of days ago.

Now I always get the following error on the last step (rendering):

Using cuda for inference. Reading video frames... Number of frames available for inference: 353 Traceback (most recent call last): File "/content/Wav2Lip/inference.py", line 280, in main() File "/content/Wav2Lip/inference.py", line 225, in main mel = audio.melspectrogram(wav) File "/content/Wav2Lip/audio.py", line 47, in melspectrogram S = _amp_to_db(_linear_to_mel(np.abs(D))) - hp.ref_level_db File "/content/Wav2Lip/audio.py", line 95, in _linear_to_mel _mel_basis = _build_mel_basis() File "/content/Wav2Lip/audio.py", line 100, in _build_mel_basis return librosa.filters.mel(hp.sample_rate, hp.n_fft, n_mels=hp.num_mels, TypeError: mel() takes 0 positional arguments but 2 positional arguments (and 3 keyword-only arguments) were given

any idea how to troubleshoot this problem? Any help would be appreciated.

Best regards!

try this colab this work https://j.mp/wav2lip

[teethdiao]

The only thing we need to do for solving this problem is to open the audio.py file and modify the funciton librosa.filters.mel in line 100 by altering "hp.sample_rate, hp.n_fft" to "sr=hp.sample_rate, n_fft= hp.n_fft".

[TheoTheGreat-stack]

The only thing we need to do for solving this problem is to open the audio.py file and modify the funciton librosa.filters.mel in line 100 by altering "hp.sample_rate, hp.n_fft" to "sr=hp.sample_rate, n_fft= hp.n_fft".

Now it just seems to get stuck on Model loaded 10% 1/10 [09:56<1:29:24, 596.06s/it]^C

But it did make more progress with your suggested changes

[TheoTheGreat-stack]

Actually all seems to be good now after a few more attempts

[11whitewater]

pip install librosa==0.8.0 maybe it can help you

[MaximillianStoner]

@TheoTheGreat-stack did you do anything additional to resolve the issue? I'm stuck in the same place you were

[TheoTheGreat-stack]

@TheoTheGreat-stack did you do anything additional to resolve the issue? I'm stuck in the same place you were

Not really, it failed once and that's when I posted on the thread, but then tried the exact same thing again and it worked. Now it seems to be fine

[dyk010518]

Hmm I am also stuck here. I have been running my model many times and it always gets stuck at the "0% 0/38 [00:00<?, ?it/s]^C"

[dyk010518]

@TheoTheGreat-stack did you do anything additional to resolve the issue? I'm stuck in the same place you were

@MaximillianStoner Did you get the issue resolved by any chance?

[sethmh82]

pip install librosa==0.8.0 This fixed it for me

[ChiragJRana]

The only thing we need to do for solving this problem is to open the audio.py file and modify the funciton librosa.filters.mel in line 100 by altering "hp.sample_rate, hp.n_fft" to "sr=hp.sample_rate, n_fft= hp.n_fft".

This worked for me. ✌️✌️

[hongxin1988]

pip install librosa==0.8.0 This fixed it for me, too. thanks!

[GregLomason]

The only thing we need to do for solving this problem is to open the audio.py file and modify the funciton librosa.filters.mel in line 100 by altering "hp.sample_rate, hp.n_fft" to "sr=hp.sample_rate, n_fft= hp.n_fft".

once following this was hit with a "NameError: name 'hp' is not defined" I'm a complete beginner at this. Smart enough to edit the code line, ignorant enough to not know what any of the language means 😂

[GregLomason]

pip install librosa==0.8.0 This fixed it for me, too. thanks!

I've installed your suggestion as well as changed the code line to "hp.sample_rate, hp.n_fft" to "sr=hp.sample_rate, n_fft= hp.n_fft". Unfortunately, I'm still being prompted with the error "NameError: name 'hp' is not defined"

@teethdiao @hongxin1988

[GregLomason]

Reference

Building wheels for collected packages: librosa Building wheel for librosa (setup.py) ... done Created wheel for librosa: filename=librosa-0.8.0-py3-none-any.whl size=201411 sha256=f3cbd8f24227ce4bf3182db8cd2b01c264a6ee4211fa413fb2aad15a0827acb4 Stored in directory: c:\users\gregl\appdata\local\pip\cache\wheels\a4\09\cc\728ed681f0fa5c37e0fbfc66d2ba07058dd995784f1f6554a8 Successfully built librosa Installing collected packages: resampy, librosa Attempting uninstall: librosa Found existing installation: librosa 0.10.0 Uninstalling librosa-0.10.0: Successfully uninstalled librosa-0.10.0 Successfully installed librosa-0.8.0 resampy-0.4.2

(voice-clone) C:\Users\gregl\Desktop\Greg's Voice Clone>python demo_cli.py Traceback (most recent call last): File "C:\Users\gregl\Desktop\Greg's Voice Clone\demo_cli.py", line 10, in from encoder import inference as encoder File "C:\Users\gregl\Desktop\Greg's Voice Clone\encoder\inference.py", line 3, in from encoder.audio import preprocess_wav # We want to expose this function from here File "C:\Users\gregl\Desktop\Greg's Voice Clone\encoder\audio.py", line 100, in sr=hp.sample_rate, n_fft= hp.n_fft NameError: name 'hp' is not defined

[RobinBrackez]

Try this:

def _build_mel_basis():
    assert hp.fmax <= hp.sample_rate // 2
    return librosa.filters.mel(sr=hp.sample_rate, n_fft=hp.n_fft, n_mels=hp.num_mels,
                               fmin=hp.fmin, fmax=hp.fmax)

Wouldn't the hp error already appear on the previous line if hp is undefined?

[pranavtushar]

Hey with the new update of librosa, you have to specify the argument name, it will work, for sample rate, sr = hp.sample_rate, and so on.

[central0658]

The only thing we need to do for solving this problem is to open the audio.py file and modify the funciton librosa.filters.mel in line 100 by altering "hp.sample_rate, hp.n_fft" to "sr=hp.sample_rate, n_fft= hp.n_fft".

I had the same issue and did this, it went on but ended in: FileNotFoundError: [Errno 2] No such file or directory: 'checkpoints/wav2lip_gan.pth'

couldn't find any fix online, any guess?

[central0658]

Using cuda for inference. Reading video frames... Number of frames available for inference: 194 (80, 575) Length of mel chunks: 176 0% 0/2 [00:00<?, ?it/s] 0% 0/11 [00:12<?, ?it/s] Recovering from OOM error; New batch size: 8

0% 0/22 [00:00<?, ?it/s] 5% 1/22 [03:45<1:18:48, 225.17s/it] 9% 2/22 [03:48<31:31, 94.56s/it]
14% 3/22 [03:51<16:42, 52.78s/it] 18% 4/22 [03:54<09:56, 33.17s/it] 23% 5/22 [03:57<06:21, 22.45s/it] 27% 6/22 [04:00<04:13, 15.85s/it] 32% 7/22 [04:03<02:54, 11.66s/it] 36% 8/22 [04:07<02:04, 8.92s/it] 41% 9/22 [04:10<01:32, 7.10s/it] 45% 10/22 [04:13<01:11, 5.95s/it] 50% 11/22 [04:16<00:55, 5.07s/it] 55% 12/22 [04:19<00:44, 4.47s/it] 59% 13/22 [04:22<00:36, 4.06s/it] 64% 14/22 [04:26<00:30, 3.86s/it] 68% 15/22 [04:29<00:25, 3.63s/it] 73% 16/22 [04:32<00:20, 3.47s/it] 77% 17/22 [04:35<00:16, 3.37s/it] 82% 18/22 [04:38<00:13, 3.38s/it] 86% 19/22 [04:42<00:09, 3.30s/it] 91% 20/22 [04:45<00:06, 3.24s/it] 95% 21/22 [04:48<00:03, 3.20s/it] 100% 22/22 [04:51<00:00, 13.25s/it] Load checkpoint from: checkpoints/wav2lip_gan.pth 0% 0/2 [05:06<?, ?it/s] Traceback (most recent call last): File "/content/Wav2Lip/inference.py", line 280, in main() File "/content/Wav2Lip/inference.py", line 252, in main model = load_model(args.checkpoint_path) File "/content/Wav2Lip/inference.py", line 171, in load_model checkpoint = _load(path) File "/content/Wav2Lip/inference.py", line 162, in _load checkpoint = torch.load(checkpoint_path) File "/usr/local/lib/python3.10/dist-packages/torch/serialization.py", line 791, in load with _open_file_like(f, 'rb') as opened_file: File "/usr/local/lib/python3.10/dist-packages/torch/serialization.py", line 271, in _open_file_like return _open_file(name_or_buffer, mode) File "/usr/local/lib/python3.10/dist-packages/torch/serialization.py", line 252, in init super().init(open(name, mode)) FileNotFoundError: [Errno 2] No such file or directory: 'checkpoints/wav2lip_gan.pth'

[G-force78]

I have this error ``` File "/content/cog-Wav2Lip/inference.py", line 15, in from batch_face import RetinaFace ModuleNotFoundError: No module named 'batch_face'

[Gumballegal]

going with librosa==0.8.0 gave me this:

PS C:\Wav2Lip-master> python inference.py --checkpoint_path wav2lip.pth --face video.mp4 --audio audio.wav Traceback (most recent call last): File "C:\Wav2Lip-master\inference.py", line 3, in import scipy, cv2, os, sys, argparse, audio File "C:\Wav2Lip-master\audio.py", line 1, in import librosa File "C:\Users\bruno\AppData\Local\Programs\Python\Python310\lib\site-packages\librosa_init_.py", line 211, in from . import core File "C:\Users\bruno\AppData\Local\Programs\Python\Python310\lib\site-packages\librosa\core_init_.py", line 9, in from .constantq import * # pylint: disable=wildcard-import File "C:\Users\bruno\AppData\Local\Programs\Python\Python310\lib\site-packages\librosa\core\constantq.py", line 1058, in dtype=np.complex, File "C:\Users\bruno\AppData\Local\Programs\Python\Python310\lib\site-packages\numpy_init_.py", line 305, in getattr raise AttributeError(former_attrs[attr]) AttributeError: module 'numpy' has no attribute 'complex'. np.complex was a deprecated alias for the builtin complex. To avoid this error in existing code, use complex by itself. Doing this will not modify any behavior and is safe. If you specifically wanted the numpy scalar type, use np.complex128 here. The aliases was originally deprecated in NumPy 1.20; for more details and guidance see the original release note at: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations. Did you mean: 'complex_'?

switching between versions of numba didn't help

[AIfrontier]

going with librosa==0.8.0 gave me this:

PS C:\Wav2Lip-master> python inference.py --checkpoint_path wav2lip.pth --face video.mp4 --audio audio.wav Traceback (most recent call last): File "C:\Wav2Lip-master\inference.py", line 3, in import scipy, cv2, os, sys, argparse, audio File "C:\Wav2Lip-master\audio.py", line 1, in import librosa File "C:\Users\bruno\AppData\Local\Programs\Python\Python310\lib\site-packages\librosa_init_.py", line 211, in from . import core File "C:\Users\bruno\AppData\Local\Programs\Python\Python310\lib\site-packages\librosa\core_init_.py", line 9, in from .constantq import * # pylint: disable=wildcard-import File "C:\Users\bruno\AppData\Local\Programs\Python\Python310\lib\site-packages\librosa\core\constantq.py", line 1058, in dtype=np.complex, File "C:\Users\bruno\AppData\Local\Programs\Python\Python310\lib\site-packages\numpy_init_.py", line 305, in getattr raise AttributeError(former_attrs[attr]) AttributeError: module 'numpy' has no attribute 'complex'. np.complex was a deprecated alias for the builtin complex. To avoid this error in existing code, use complex by itself. Doing this will not modify any behavior and is safe. If you specifically wanted the numpy scalar type, use np.complex128 here. The aliases was originally deprecated in NumPy 1.20; for more details and guidance see the original release note at: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations. Did you mean: 'complex_'?

[RobinBrackez]

@Gumballegal @central0658 The checkpoints/wav2lip_gan.pth has to be downloaded separately, the links are in the README (topic: "Getting the weights"). So if the script says the file is missing, you have to download it and put it in that checkpoints-folder.

@AIfrontier The Numpy error makes me think you have the wrong version installed.

These are the versions I use in my venv. Numpy is 1.24.3

librosa               0.10.0.post2
numpy                 1.24.3
opencv-python         4.7.0.72
threadpoolctl         3.1.0
torch                 1.11.0+cu113
torchvision           0.12.0+cu113

[G-force78]

This one works fine https://colab.research.google.com/drive/1OlQxvo5IX4zCpE9UWTL_vkRzpu5nqQoM?usp=sharing

[PrabalS12]

This one works fine https://colab.research.google.com/drive/1OlQxvo5IX4zCpE9UWTL_vkRzpu5nqQoM?usp=sharing

gives me the same error during inference phase:

Using cuda for inference. Reading video frames... Number of frames available for inference: 400 (80, 641) Length of mel chunks: 393 0% 0/4 [00:00<?, ?it/s] 0% 0/25 [00:00<?, ?it/s]^C

Neither did the modifications in audio.py help nor did the pip install librose==0.8.0 help

Stuck here --> Using cuda for inference. Reading video frames... Number of frames available for inference: 400 (80, 641) Length of mel chunks: 393 0% 0/4 [00:00<?, ?it/s] 0% 0/25 [00:00<?, ?it/s]^C

[G-force78]

Thats odd its probably the size of the image files you are using, make them square or divisible by 16

GitHub Copilot: No, not all video resolutions are divisible by 16. However, many common video resolutions are divisible by 16, such as 1080p (1920x1080), 720p (1280x720), 480p (854x480), and 360p (640x360). This is because video resolutions are often designed to have a width and height that are multiples of 16, which can make it easier to compress and decompress the video data. However, there are also many video resolutions that are not divisible by 16, such as 4K (3840x2160) and 8K (7680x4320).

[gretoverno]

The only thing we need to do for solving this problem is to open the audio.py file and modify the funciton librosa.filters.mel in line 100 by altering "hp.sample_rate, hp.n_fft" to "sr=hp.sample_rate, n_fft= hp.n_fft".

I think this plus just removing all the version requirements in the requirements.py file works for me

[ankitvirla]

pip install librosa==0.8.0 maybe it can help you

thanks, it really worked :)

[Jerry-Master]

for librosa==0.10.1 you can fix this error by changing line 100 of audio.py to

return librosa.filters.mel(sr=hp.sample_rate, n_fft=hp.n_fft, n_mels=hp.num_mels,
                               fmin=hp.fmin, fmax=hp.fmax)

You just need to specify the sr= and n_fft= keywords. Apparently they moved them from being positional to being keyword.