Reference voice

Record a sample

3–30 seconds · quiet room · one speaker

00:00

Upload WAV, MP3, M4A, or WebM

Reference sample Ready

Exact transcript Improves similarity

Skip the transcript and use speaker-only mode (lower quality)

Generate speech

Text to speak 0/800

Language 10 languages supported

Generating your audio Connecting to ZeroGPU…

Your audio is ready Generated with Qwen3-TTS