Skip to main content
Seed-Audio 1.0 selects a voice through audio_references in the audio generation API. Each entry can be:
  • Preset voice ID — use a voice_type from the table below, e.g. zh_female_vv_uranus_bigtts
  • Reference audio URL — upload a reference clip for voice cloning
All preset voices support controlling emotion, tone, and style through natural-language prompts. Chinese voices can also read English text.
In prompt, use @audioN to reference the Nth entry in audio_references (numbering starts at 1), letting you mix multiple voices in one clip. The “2.0” in a voice name marks the voice-library version and is unrelated to the model version.

General Voices

288 voices in total — mostly Chinese, plus 15 ICL character voices (American / Australian / British English).