[GH-ISSUE #1904] How to train a high-quality model #1281

New issue

Closed

opened 2026-10-20 10:07:38 +00:00 by bsakaguchi · 2 comments

bsakaguchi commented

2026-10-20 10:07:38 +00:00

Owner

Originally created by @sunnyhui on GitHub (Mar 21, 2024).
Original GitHub issue: https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/issues/1904

I am currently attempting to train a base model, but it is difficult to achieve a level of similarity comparable to that of existing open-source models. I was wondering if there are any specific techniques or strategies for effectively training a base model. I would greatly appreciate a detailed response.

Originally created by @sunnyhui on GitHub (Mar 21, 2024). Original GitHub issue: https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/issues/1904 I am currently attempting to train a base model, but it is difficult to achieve a level of similarity comparable to that of existing open-source models. I was wondering if there are any specific techniques or strategies for effectively training a base model. I would greatly appreciate a detailed response.

bsakaguchi

2026-10-20 10:07:38 +00:00

closed this issue
added the
stale
label

bsakaguchi commented

2026-10-20 10:07:38 +00:00

Author

Owner

@github-actions[bot] commented on GitHub (May 6, 2024):

This issue was closed because it has been inactive for 15 days since being marked as stale.

@github-actions[bot] commented on GitHub (May 6, 2024): This issue was closed because it has been inactive for 15 days since being marked as stale.

bsakaguchi commented

2026-10-20 10:07:39 +00:00

Author

Owner

@jupinter commented on GitHub (Jun 17, 2024):

I am currently attempting to train a base model, but it is difficult to achieve a level of similarity comparable to that of existing open-source models. I was wondering if there are any specific techniques or strategies for effectively training a base model. I would greatly appreciate a detailed response.

我最近也在做类似的事情，用现有的预训练模型在某些情况下存在发音异常，所以也希望重训一个更好的模型。但是训了一段时间之后效果仍不理想。

@jupinter commented on GitHub (Jun 17, 2024): > I am currently attempting to train a base model, but it is difficult to achieve a level of similarity comparable to that of existing open-source models. I was wondering if there are any specific techniques or strategies for effectively training a base model. I would greatly appreciate a detailed response. 我最近也在做类似的事情，用现有的预训练模型在某些情况下存在发音异常，所以也希望重训一个更好的模型。但是训了一段时间之后效果仍不理想。