[GH-ISSUE #984] After converting audio twice in a row, the second output has a much deeper pitch? #642

New issue

Closed

opened 2026-10-20 10:01:31 +00:00 by bsakaguchi · 3 comments

bsakaguchi commented

2026-10-20 10:01:31 +00:00

Owner

Originally created by @moolzebub on GitHub (Aug 12, 2023).
Original GitHub issue: https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/issues/984

I have NO CLUE how to approach explaining this issue, but it's consistently happening.
I do the regular steps towards converting audio and it goes without a hitch the first time! Then I convert the same audio file cause I changed some settings, but the result is much deeper and noticeably slower than the previous.
I found a way to temporarily stop it from doing this: just unloading the voice. Okay, fine, I can deal with unloading it everytime... but when it comes to batch conversion it's annoying!

The first audio generated is fine.
Then the next generation, no different settings, no unloading.

Originally created by @moolzebub on GitHub (Aug 12, 2023). Original GitHub issue: https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/issues/984 _I have NO CLUE how to approach explaining this issue, but it's consistently happening._ I do the regular steps towards converting audio and it goes without a hitch the first time! Then I convert the same audio file cause I changed some settings, but the result is much **deeper** and noticeably **slower** than the previous. I found a way to temporarily stop it from doing this: just unloading the voice. Okay, fine, I can deal with unloading it everytime... but when it comes to batch conversion it's annoying! [The first audio generated is fine.](https://drive.google.com/file/d/1O26BTg-so0AFx7wqokQq8xXBenHkAFc1/view?usp=drive_link) [Then the next generation, no different settings, no unloading.](https://drive.google.com/file/d/1NgR762r11pR7ic7qdrWF9LtGnEc1HFpD/view?usp=drive_link)

bsakaguchi closed this issue

2026-10-20 10:01:31 +00:00

bsakaguchi commented

2026-10-20 10:01:31 +00:00

Author

Owner

@ben-mkiv commented on GitHub (Aug 13, 2023):

also happens in the opposite direction, with female voices the pitch increases.

maybe the pitch algorithm result doesn't get purged correctly?!

same issue: https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/issues/885

@ben-mkiv commented on GitHub (Aug 13, 2023): also happens in the opposite direction, with female voices the pitch increases. maybe the pitch algorithm result doesn't get purged correctly?! same issue: https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/issues/885

bsakaguchi commented

2026-10-20 10:01:31 +00:00

Author

Owner

@JarodMica commented on GitHub (Aug 14, 2023):

The issue occurs when resampling the audio. If you leave it at 0, this does not occur, though, it's definitely a bug. It's been in RVC since I've been using it back for a few months now, though, I don't notice much of a difference between audio at 48k vs 0 so I just leave it there.

@JarodMica commented on GitHub (Aug 14, 2023): The issue occurs when resampling the audio. If you leave it at 0, this does not occur, though, it's definitely a bug. It's been in RVC since I've been using it back for a few months now, though, I don't notice much of a difference between audio at 48k vs 0 so I just leave it there.

bsakaguchi commented

2026-10-20 10:01:31 +00:00

Author

Owner

@RVC-Boss commented on GitHub (Aug 16, 2023):

Fixed now.
github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI@deb756bf1b

@RVC-Boss commented on GitHub (Aug 16, 2023): Fixed now. https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/commit/deb756bf1b0bc5924865e646dc7925c68c9f5bc5