[GH-ISSUE #720] Regarding inputted speaker content #441

New issue

Closed

opened 2026-10-20 09:59:22 +00:00 by bsakaguchi · 1 comment

bsakaguchi commented

2026-10-20 09:59:22 +00:00

Owner

Originally created by @suzhenghang on GitHub (Jul 10, 2023).
Original GitHub issue: https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/issues/720

Because retrieval features are used during inference to replace input features in order to prevent speaker identity leakage, but how can we ensure that the generated speech still corresponds to the original input content in terms of the speaker's voice?

feats = (
torch.from_numpy(npy).unsqueeze(0).to(self.device) * index_rate
+ (1 - index_rate) * feats
)

Originally created by @suzhenghang on GitHub (Jul 10, 2023). Original GitHub issue: https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/issues/720 Because retrieval features are used during inference to replace input features in order to prevent speaker identity leakage, but how can we ensure that the generated speech still corresponds to the original input content in terms of the speaker's voice? feats = ( torch.from_numpy(npy).unsqueeze(0).to(self.device) * index_rate + (1 - index_rate) * feats )

bsakaguchi closed this issue

2026-10-20 09:59:22 +00:00

bsakaguchi commented

2026-10-20 09:59:23 +00:00

Author

Owner

@RVC-Boss commented on GitHub (Jul 11, 2023):

Replace with the most similar features to ensure semantic consistency.

@RVC-Boss commented on GitHub (Jul 11, 2023): Replace with the most similar features to ensure semantic consistency.

bsakaguchi referenced this issue

2026-10-20 10:02:34 +00:00

[GH-ISSUE #1123] RVC colab disconnect #744