I keep getting this error "ValueError" when trying to create

[AKanoniOS]

ValueError: can't extend empty axis 0 using modes other than 'constant' or 'empty'

Does anyone know why this is happening? Thanks

[AKanoniOS]

Here is the full text of what happens when I try to preview or create

(mugen) steve@steves-MacBook-Pro mugen % mugen preview -a ../alteredsnaps-OriginalAudio.mp3

Analyzing audio...
Traceback (most recent call last):
File "/Users/steve/opt/miniconda3/envs/mugen/bin/mugen", line 8, in
sys.exit(main())
File "/Users/steve/opt/miniconda3/envs/mugen/lib/python3.9/site-packages/scripts/cli/cli.py", line 27, in main
args.func(args)
File "/Users/steve/opt/miniconda3/envs/mugen/lib/python3.9/site-packages/scripts/cli/commands.py", line 144, in preview_music_video
events = prepare_events(generator, args)
File "/Users/steve/opt/miniconda3/envs/mugen/lib/python3.9/site-packages/scripts/cli/events.py", line 56, in prepare_events
events = get_events_from_audio(generator.audio, args)
File "/Users/steve/opt/miniconda3/envs/mugen/lib/python3.9/site-packages/scripts/cli/events.py", line 78, in get_events_from_audio
events = get_beat_events(audio, beats_mode)
File "/Users/steve/opt/miniconda3/envs/mugen/lib/python3.9/site-packages/scripts/cli/events.py", line 89, in get_beat_events
events = audio.beats()
File "/Users/steve/opt/miniconda3/envs/mugen/lib/python3.9/site-packages/mugen/audio/Audio.py", line 105, in beats
untrimmed_beats = self._beats()
File "/Users/steve/opt/miniconda3/envs/mugen/lib/python3.9/site-packages/mugen/audio/Audio.py", line 155, in _beats
tempo, beats = librosa.beat.beat_track(
File "/Users/steve/opt/miniconda3/envs/mugen/lib/python3.9/site-packages/librosa/beat.py", line 179, in beat_track
onset_envelope = onset.onset_strength(
File "/Users/steve/opt/miniconda3/envs/mugen/lib/python3.9/site-packages/librosa/onset.py", line 333, in onset_strength
odf_all = onset_strength_multi(
File "/Users/steve/opt/miniconda3/envs/mugen/lib/python3.9/site-packages/librosa/onset.py", line 568, in onset_strength_multi
S = np.abs(feature(y=y, sr=sr, n_fft=n_fft, hop_length=hop_length, **kwargs))
File "/Users/steve/opt/miniconda3/envs/mugen/lib/python3.9/site-packages/librosa/feature/spectral.py", line 1995, in melspectrogram
S, n_fft = _spectrogram(
File "/Users/steve/opt/miniconda3/envs/mugen/lib/python3.9/site-packages/librosa/core/spectrum.py", line 2512, in _spectrogram
stft(
File "/Users/steve/opt/miniconda3/envs/mugen/lib/python3.9/site-packages/librosa/core/spectrum.py", line 228, in stft
y = np.pad(y, int(n_fft // 2), mode=pad_mode)
File "<array_function internals>", line 180, in pad
File "/Users/steve/opt/miniconda3/envs/mugen/lib/python3.9/site-packages/numpy/lib/arraypad.py", line 814, in pad
raise ValueError(
ValueError: can't extend empty axis 0 using modes other than 'constant' or 'empty'

[AKanoniOS]

I solved it, the audio I was using was broken or something, or Mugen doesn't like mp3. I converted it to .wav and now it works!

[scherroman]

Unsure what the issue was in this case, though I've never run into any issues with any mp3s personally, and they're also used in the end-to-end test suite / tested cross-platform. Note that under the hood mugen uses librosa to analyze the audio, which shouldn't have issues reading mp3s as long as ffmpeg is installed, which is installed when installing mugen