seed-icl-2.0Audio

Doubao Seed ICL 2.0

Doubao Voice Cloning 2.0: clone a voice from your own recording and synthesize with it, official price ¥3 per 10,000 characters.

Notes

seed-tts-2.0: without voice, or with an OpenAI voice name, the default voice follows the text's language: zh_female_vv_uranus_bigtts for Chinese, en_female_dacey_uranus_bigtts otherwise. seed-audio-1.0: write "scene description + lines" in input; without voice, or with an OpenAI voice name, the description decides the voice.
seed-tts-2.0 takes about 1–3 s for a short sentence and about 70 s for 2,000 characters; seed-audio-1.0 takes about 10–30 s.
Cloning takes about half a minute (a 17-second recording took about 33 s in our test); use an HTTP read timeout of at least 300 s. When cloned-voice capacity is full the request returns 503 voice_capacity_full; try again later.

Endpoints