[GH-ISSUE #2029] Giving my voice model an accent #1371

New issue

Open

opened 2026-10-20 10:08:31 +00:00 by bsakaguchi · 2 comments

bsakaguchi commented

2026-10-20 10:08:31 +00:00

Owner

Originally created by @EldritchSpiral on GitHub (Apr 30, 2024).
Original GitHub issue: https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/issues/2029

So I'm working on a novel where one of the main characters is a singer and I decided to create a voice model for her so I can hear her sing in her actual voice. I actually have several "voice claims" I want to use, but I have a problem when it comes to her accent, since none of the voice claims have her nationality, and hence, not the accent I'm looking for. So essentially I'm curious is there any way to give my model a specific accent without changing her voice much?

Originally created by @EldritchSpiral on GitHub (Apr 30, 2024). Original GitHub issue: https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/issues/2029 So I'm working on a novel where one of the main characters is a singer and I decided to create a voice model for her so I can hear her sing in her actual voice. I actually have several "voice claims" I want to use, but I have a problem when it comes to her accent, since none of the voice claims have her nationality, and hence, not the accent I'm looking for. So essentially I'm curious is there any way to give my model a specific accent without changing her voice much?

bsakaguchi added the

enhancement

label

2026-10-20 10:08:31 +00:00

bsakaguchi commented

2026-10-20 10:08:32 +00:00

Author

Owner

@78Alpha commented on GitHub (May 4, 2024):

If setting index ratio to 1 doesn't add enough accent, then you'd need someone already singing with the accent. RVC layers a voice on top, it didn't usually do such drastic alterations to the source.

@78Alpha commented on GitHub (May 4, 2024): If setting index ratio to 1 doesn't add enough accent, then you'd need someone already singing with the accent. RVC layers a voice on top, it didn't usually do such drastic alterations to the source.

bsakaguchi commented

2026-10-20 10:08:32 +00:00

Author

Owner

@EldritchSpiral commented on GitHub (May 6, 2024):

If setting index ratio to 1 doesn't add enough accent, then you'd need someone already singing with the accent. RVC layers a voice on top, it didn't usually do such drastic alterations to the source.

The thing is that for index ratio to work I need my model to have the accent already, which is exactly the problem since none of the voice claims I have for her have the accent I need. Would merging my result model with another voice with the accent be enough? What's the minimum weight a model should have in the merging process to pick up its accent while having mostly the timbre of the other model's voice if it's possible?

@EldritchSpiral commented on GitHub (May 6, 2024): > If setting index ratio to 1 doesn't add enough accent, then you'd need someone already singing with the accent. RVC layers a voice on top, it didn't usually do such drastic alterations to the source. The thing is that for index ratio to work I need my model to have the accent already, which is exactly the problem since none of the voice claims I have for her have the accent I need. Would merging my result model with another voice with the accent be enough? What's the minimum weight a model should have in the merging process to pick up its accent while having mostly the timbre of the other model's voice if it's possible?