Real-Time-Voice-Cloning
The classic 5-second voice cloning project — the "founding father" of voice cloning
The classic voice-cloning project that clones voices from 5 seconds of audio — the "founding father" of the voice cloning scene.
Text-to-speech, speech-to-text, and voice cloning, open source.
Key features
- Three-stage SV2TTS: speaker encoder, Tacotron synthesizer, WaveRNN vocoder
- Clones a voice from a few seconds of audio
- Real-time vocoder for live synthesis
- Interactive toolbox GUI plus CLI mode; pretrained models auto-download
Good for
- Researchers reproducing the SV2TTS speaker-verification-to-TTS paper
- Hobbyists experimenting with voice cloning on CPU or GPU
Stack & tags
Pythonvoice-cloningttssv2ttsresearchreal-timetoolbox-gui
Quick start
uv run --extra cpu demo_toolbox.pyFrom the project README on GitHub.
Screenshots
Screenshots from the project’s GitHub README.
Similar open-source projects
Whisper
★ 109,795OpenAI's open-source speech recognition model — the benchmark for multilingual STT
GPT-SoVITS
★ 62,271TTS/cloning trainable on 1 minute of audio — a legendary Chinese-community project