[2025/12/01] Authorized for commercial use: Wan Sheng Shaoyu Yin RVC model
Inference Comparison
After Inference
Other Inference
Model Introduction
Please note that version V3 is Cn_hubert version, so strictly follow the https://klrvc.com/zh/tutorial/3714 tutorial to deploy the model.
Last updated: 2025-12-01
This update strengthens articulation by adopting Cn_hubert self-supervision features to enhance articulation and speech confusion.
Note that the V3 version requires replacing the Cn_hubert and is not compatible with the V1 and V2 versions. Please follow the instructions above to replace the one.
Updated: 2025-04-07
This update introduces the second version, with minor adjustments to the first model, adding emotional expression, better clarity, and reduced some reverberation.
Below is the training diagram for the second version of this update. After 10K, the model began to overfit, and it was decided to use the 10.2K step model as the inference model.
![Image [1]-Miaoyin-RVC Tone Model Workshop](https://apis.klrvc.com/wp-content/uploads/2024/06/07a1f63c5420250407152503.png)
—–Divider ———-
This is Miaoyin's first collaboration with top streamer Wan Sheng. In the early stages, we recorded a large amount of daily chat, livestream chat, and singing segments, with total training data lasting over an hour. We proportionally integrated daily chatting and singing, ultimately producing a very high-quality model. We also handled pitch and bass very well. While we wouldn't call it a full-range model, it was almost a full-range model. Wansheng has also partnered with Miaoyin. After purchase, the RVC model automatically receives authorization and can be used commercially (meaning you can livestream or anything else) but must not be used for purposes that violate laws and regulations. Otherwise, it will bear the corresponding legal risks yourself.
Next, I will show the dry notes and a cappella singing effects after reasoning, which better reflect the model's quality.
No reasoning has been
made <!--[if lt IE 9]>document.createElement('audio');<![endif]-->After reasoning
A cappella and a cappella reasoning
All the above audio test results come from Miaoyin Workshop reasoning, with the following parameters
Parameters
Tone change: 0
Algorithm: rmvpe
Training: 500
Sampling rate: 44K
Project documents
![image[2]-Miaoyin-RVC Sound Model Workshop](https://apis.klrvc.com/wp-content/uploads/2024/06/20240611145654790-WX20240611-145715-300x150.png)
Others
Although this is an authorized commercial version, please do not use it for illegal purposes, or you may face legal consequences. Also, please do not re-share the information after purchase. The model contains basic information of member buyers. If such redistribution is discovered, we will suspend your usage rights. Thank you for your understanding. Miaoyin Workshop.
Version Details
Recommended Parameters
Related Models
Keywords
Comments
No comments yet

![[2025/12/01] Authorized for commercial use: Wan Sheng Shaoyu Yin RVC model](https://apis.klrvc.com/wp-content/uploads/2024/06/2cbb54265020250617163854.png)
