RVC2.3.260718 released

An Important Update After two years, RVC-Boss brings a significant update to his proud project, Retrieval-based-Voice-Conversion-WebUI. The updates are as follows: Reduced webpage...

An Important Update After two years, RVC-Boss brings a significant update to his proud project, Retrieval-based-Voice-Conversion-WebUI. The updates are as follows: Reduced webpage usage difficulty (inference input audio requires no input path, frontend real-time log optimization, automatic refreshing of model lists after saving inference small models, automatic detection of matching index paths after selecting models on the inference homepage, etc.) Data feature extraction, model training, and indexing stages can be interrupted midway and resume progress restarted. GPU automatically recognizes and judges inference patterns and accuracy. Fixes dependency issues (e.g., pyworld, fairseq, matplotlib, pyav, onnxruntime-GPU environment repair, etc.) Solving coding issues, automatically searching for available ports when launching a webpage, single-card training without DDP The base model remains unchanged. Fixing some effects can be improved by updating the integration package/code package version with existing fine-tuning models, or retraining and inference with a new version. Supports FCPE pitch extraction algorithm. Real-time voice-changing mode GUI supports cuda graph inference acceleration, with algorithm inference delay indicators speeding up to 4.7 times (4090D test results), while graphics card usage is reduced by about 20~30%. Obviously obsolete or outdated features have been removed. The backend for separating vocals and accompaniment has been replaced from UVR5 with PyMSS, and inference interruptions are supported (a better vocal separation algorithm can significantly improve inference performance). Optimized input audio reading and resampling processes. Use GPU processing when conditions permit to improve inference efficiency and reduce CPU usage. Fix libuv compatibility issues for training processes in Windows environments. Download: Complete package For Nvidia GPU users (<50x0): https://huggingface.co/lj1995/VoiceConversionWebUI/resolve/main/RVC20260718Nvidia.7z For Nvidia GPU users (>=50x0): https://huggingface.co/lj1995/VoiceConversionWebUI/resolve/main/RVC20260718Nvidia50x0.7z For AMD/Intel GPU users: https://huggingface.co/lj1995/VoiceConversionWebUI/resolve/main/RVC20260718AMD_Intel.7z RVC Real-Time VST2/VST3 Plugin (For user instructions, please refer to the README included in the archive. You should unzip the complete package first.) https://huggingface.co/lj1995/VoiceConversionWebUI/resolve/main/RVCRealtime-Win64VST.zip Chinese users can use the following two sources to accelerate downloads: 1. Free full-speed download of direct link collections: https://www.yuque.com/flowercry/hxf0ds 2. Dupan requires a Super Member to download without speed limits. Link: https://pan.baidu.com/s/16BHeTXRq2aGUzDZo1BY1Lw Extraction code: tvy6

MiaoYin Original Content. Official source: https://klrvc.com. Source: https://klrvc.com/en/updates/13. Unauthorized scraping, republishing, model data cloning, or commercial redistribution is not permitted.