You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I finetuned using 40 mins of audio of Urdu for more than 100 epochs but the voice I get is not that clean. How much data do I need to get good results or do I need to update or change anything else.
The text was updated successfully, but these errors were encountered:
Make sure your audios don't have noise. Generate a spectrogram to visualize noise. Try increasing the sampling rate to 44100, if you can handle the computation.
I finetuned using 40 mins of audio of Urdu for more than 100 epochs but the voice I get is not that clean. How much data do I need to get good results or do I need to update or change anything else.
The text was updated successfully, but these errors were encountered: