Retrieval-based-Voice-Conversion-WebUI
Retrieval-based-Voice-Conversion-WebUI copied to clipboard
Cracking sound when separating vocals and instrumentals (with back vocals)
Hi! I have a weird issue where I always get some cracking sound every few seconds after separating vocals and instrumental + vocals. Cracking is present on both files. For the track in the attachment, you can start hearing it after 00:16. To me it sounds like bass is making a problem but I am not certain.
Exactly the same problem. when using HPx_.... models. When percussion instruments (mostly kicks) are played, a sound appears on the instrumental track that reaches the volume limit, resulting in a “crackling” sound.
Because of this problem, I went to try third-party models, but ran into a problem that I described here -> #1765
If you find a model that resolves this please let me know.
I couldn't find the model, but I was able to solve the problem with the cracks. So it turned out that the audio could have a high volume, and uvr could actually increase the volume. The result was a volume higher than maximum => crackling
In short, try reducing the volume of the audio input and the problems should go away
I have this issue with HP5_only_main_vocal
model. This is how the resulting vocals look in audacity:
Right channel is totally f*cked, not only crackling
Use noise reduction techniques to remove artifacts.
import noisereduce as nr
y, sr = librosa.load('your_audio_file.wav', sr=None) # Load your audio file
y_reduced = nr.reduce_noise(y=y, sr=sr) # Apply noise reduction
librosa.output.write_wav('reduce_noise.wav', y_reduced, sr) # Save the noise-reduced audio