fishaudio / fish-diffusion

An easy to understand TTS / SVS / SVC framework
https://diff.fish.audio
MIT License
635 stars 81 forks source link

Dealing with Scratchy Output and Octave Shift? #81

Closed chigkim closed 1 year ago

chigkim commented 1 year ago

I have two problems that I often encounter, and I wonder if anyone has suggestion.

  1. Output audio with scratchy parts
  2. Some pitches get shifted either one octave lower or higher than the source

I'm attaching a sample output only highlighting the problems. Thanks! Scratchy.zip

leng-yue commented 1 year ago

Which pitch extractor are u using?

chigkim commented 1 year ago

I'm just using the default configuration from svc_hifisinger_finetune.py.

    pitch_extractor=dict(
        type="ParselMouthPitchExtractor",
        keep_zeros=False,
        f0_min=40.0,
        f0_max=1600.0,
    ),
leng-yue commented 1 year ago

Try CrepePitchExtractor when your input quality is not that good.