Text to speech

Relyven uses Doubao TTS 2.0 to synthesize multilingual speech.

Capabilities

Model	Capabilities	Output and parameters
doubao-tts-2.0	Text to speech in 16 languages	mp3, wav, or opus; speed from 0.5 through 2.0

Supported languages

Chinese, English, Japanese, Korean, Mexican Spanish, Brazilian Portuguese, Indonesian, Malay, Thai, Vietnamese, Filipino, German, French, Spanish (Spain), Russian, and Arabic.

Endpoint

POST https://api.relyven.com/v1/audio/speech

The request requires model, input, and voice.

Example voices

The table lists the voices used in these examples, not the complete voice catalog. See the official Volcengine voice list for other languages and voices.

Language	Voice ID
Chinese	zh_female_vv_uranus_bigtts
English	en_female_jane_uranus_bigtts

curl https://api.relyven.com/v1/audio/speech \
  -H "Authorization: Bearer $RELYVEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "doubao-tts-2.0",
    "input": "Welcome to Relyven.",
    "voice": "en_female_jane_uranus_bigtts",
    "response_format": "mp3",
    "speed": 1
  }' \
  --output speech.mp3

Check the HTTP status, file signature, and duration; do not rely on the file extension alone. Do not put keys or private input text in reports.