TTS

text to speech

Type something and synthesise it in the reference voice. The clip is shared with the voice page — change it under Reference voice.

Temperature 0.90Top-p 1.00Repetition penalty 1.05Top-k 50

No reference voice

Transcript of the clip — this is sent as ref_text and strongly affects clone quality. It must match what is actually said.

Choose clip

This TTS server clones from a clip sent with every sentence, so the call speaks in whatever voice is here. 5–15 seconds of clean speech works best. Stored in this browser only.