Skip to content

Text to speech ​

Synthesize MP3 with Speech Light, Speech Pro, or Speech Expressive. Read GET /v1/speech/catalog for available voices and local prices. Legacy Qwen voices are no longer supported.

http
POST /v1/chat/speech
ParametersTypeRequiredDescription
textstringYesText to be read; maximum 2,000 UTF-16 units
voicestringNoComplete catalog voice ID, such as speech-light:af_alloy; defaults to an available configured voice
streambooleanNoDefault false; true returns binary MP3 audio, not JSON or SSE
bash
curl -X POST https://api.modelgo.moneygo.ai/v1/chat/speech \
  -H "Authorization: Bearer sk-your-key" \
  -H "Content-Type: application/json" \
  -d '{"text":"The API integration is working","voice":"speech-light:af_alloy"}'

The non-streaming response includes audio (data:audio/mpeg;base64,...), voice, and billing with local credits, billable characters, model and price per 1,000 characters. Audio can be used as an <audio> element's src.

Each model is priced separately per 1,000 UTF-16 units, using upstream's normalized character count and rounding the total up to eight decimals. Ordinary letters count as one unit; 😀 counts as two. Failed synthesis releases reserved credits. Replaying existing audio is free; synthesizing again incurs a new charge.

For OpenAI SDKs use POST /v1/audio/speech with model, input, voice and optional response_format="mp3"; it returns binary MP3. Only this documented parameter subset is supported. See full examples, streaming, billing and error handling.

Copyright © 2026 ModelGo all rights reserved | Technology powered by MoneyGo.ai