Retrieval-based Voice Conversion WebUI

Use it when

• You have recorded vocals or speech and want them re-voiced in another timbre while keeping the melody, words, and timing
• You need a near realtime voice changer for live use, with documented end-to-end latency around 170 ms (about 90 ms with ASIO devices)

What it solves

Not the fit when

• speech-to-text transcription
• text-to-speech synthesis without a source performance to convert
• Unauthorized impersonation, fraud, or use of a target voice without the speaker's consent and appropriate rights
• media-transcript-quality
• ocr-needed

Install

python3.12 -m venv .venv && source .venv/bin/activate && python -m pip install -r requirments_cpu_py312.txt (or requirments_cu118_py312.txt / requirments_cu128_py312.txt after installing matching torch); then hf download lj1995/VoiceConversionWebUI --include "hubert_base/*" --local-dir assets and hf download lj1995/VoiceConversionWebUI rmvpe.pt --local-dir assets/rmvpe

Invoke

python webui.py (WebUI on port 7865; --noautoopen on headless servers); put user .pth models in assets/weights/ and .index files in assets/indices/; go-realtime_gui.bat for the realtime voice changer on Windows

Alternatives

No reviewed alternatives recorded yet.