RIN_E3 (rvc) Pretrain!! Download link - https://huggingface.co/MUSTAR/RIN_E3/tree/main Pretrain "RIN_E3" improves quality for English models. Use cases if: dataset is clean. Pros: better in terms of quality (use clean dataset!) Less dataset needed (recommend 1m and more) Cons: really sensitive to noise and audio distortion How to use: 1 Download G and D pth files from Huggins 2 Place it in the pretrained_v2\custom pretrained folder if you have Applio, upload it directly (but you can also send it to the folder) 3 Paste the path to the pretrain in the fork (standard fork or mangio fork) 4 You're ready to Rock! dataset: 140hrs epochs: 50 steps: 1040928 sample rate: 40k hardware trained on: rtx 4080, a100>  - **Developed by:** [MUSTAR]