MiniMax

Route to MiniMax by prefixing the model with minimax/. speech-2.8-hd is the high-fidelity model; speech-2.8-turbo trades some quality for lower latency and cost.

Models

Streaming isn't available: the gateway returns the complete audio once it's generated. Word timestamps on the /v1/audio/speech/with-timestamps endpoint come from the gateway's timestamp fallback. Requests over the 3000 character input limit are chunked and stitched server-side.

Usage

curl -X POST https://api.speechbase.ai/v1/audio/speech \
  -H "Authorization: Bearer $SPEECHBASE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "mode": "inline",
    "model": "minimax/speech-2.8-hd",
    "voice": "Wise_Woman",
    "text": "Hello from Speechbase!",
    "output": "mp3"
  }' --output hello.mp3

Replace voice with a MiniMax voice ID, or register one as a saved Voice.

Provider options

Anything in providerOptions is forwarded to the MiniMax API unchanged.