🗣️ Audio8 TTS Preview 0.6B

Multilingual text-to-speech with zero-shot voice cloning — a 0.6B DualAR model plus a 44.1 kHz neural codec, ranking in the first tier of SOTA TTS systems. Give it text, a short reference clip and that clip's transcript.

Model · GitHub · Audio samples

Text to speak

Reference voice (optional)

Reference transcript

Exactly what is said in the reference clip.

Generated speech

Pick a built-in voice below to load its reference clip and transcript, or upload your own and type exactly what it says. Built-in voices are the authors' own examples from their Apache-2.0 Audio8 TTS Preview 0.6B Space. Only clone a voice you have permission to use, and disclose synthetic audio.

Examples