RVC WebUI
github.com
- Free plan
- Windows
- Linux
- Self-hosted
What RVC WebUI is
RVC WebUI is free, MIT-licensed software for converting voice timbre and changing voices. It runs as self-hosted desktop software on Linux and Windows, with support for recorded-audio processing, audio export, and real-time changing. The project says a model can be trained from a small amount of speech and recommends at least 10 minutes of low-noise recording. Top-1 retrieval substitutes source features with training-set features to reduce tone leakage. Checkpoint processing can merge models to alter timbre. The interface can separate vocals and instruments with pymss/MSST models and includes the RMVPE high-pitch voice extraction algorithm. Reported real-time latency is 170 ms, or 90 ms with ASIO devices when drivers support it. Setup requires local Python, model downloads from Hugging Face, and suitable hardware.
Who it suits
It suits users who can host desktop software and configure Python, models, and audio hardware. It is relevant to people changing recorded or live voice audio on Linux or Windows.
Good to know, and what to watch
- Free MIT-licensed software
- Recorded and real-time voice changing
- Model training from limited speech
- Vocal and instrument separation
- Audio export supported
- Requires local Python and model setup
- Model files must be downloaded separately
- GPUs under 4 GB VRAM may be unusable
- CUDA memory errors may require smaller batches
Plans
Every plan and priceRVC WebUI
Free plan
The verdict
RVC WebUI combines voice conversion, model training, retrieval, separation, and real-time processing in a self-hosted package. Its setup and hardware requirements matter before choosing it.
What the maker publishes
- GPU support
- The guide lists CPU/AMD/Intel dependencies and separate CUDA 12.8 and CUDA 11.8 installation paths for NVIDIA GPUs. source
- Integrations
- The documented project components and credits include Gradio, FFmpeg, Ultimate Vocal Remover, pymss, and audio-slicer. source
- License
- The repository lists the project under the MIT license. source
- Model fusion
- A checkpoint processing option can merge models to change timbre. source
- Model setup
- The guide requires users to download model files from a Hugging Face model repository for inference and feature extraction. source
- Operating systems
- The setup guide covers Windows and Ubuntu 24.04 x86_64, and says Linux uses CPU for AMD/Intel systems while Windows may use DirectML. source
- Pitch extraction
- The project includes the RMVPE high-pitch voice extraction algorithm and describes it as fast with low resource consumption. source
- Purpose
- RVC WebUI is a voice timbre conversion and voice changer framework. source
- Real-time conversion
- The project reports 170 ms end-to-end latency, or 90 ms with ASIO input and output devices depending on hardware driver support. source
- Resource limits
- The FAQ says GPUs with less than 4 GB of VRAM may be unusable for the documented workflow and explains that CUDA out-of-memory can require reducing batch size. source
- Retrieval
- Top-1 retrieval replaces source features with training-set features to reduce tone leakage. source
- Training
- The project describes training as fast, including on relatively weak graphics cards. source
- Training data
- The project says it can train a voice conversion model with a small amount of data and recommends at least 10 minutes of low-noise speech. source
- Vocal separation
- The interface can use pymss/MSST models to separate vocals and instruments. source
Checked against the maker's pages on 9 Oct 2026.