Global emotion
Set one emotion for an entire text-to-speech request.
Global emotion
The global emotion field adjusts how an entire text-to-speech request is delivered without changing the selected voice.
Global emotion
{
"text": "We finally finished the project. This is wonderful news.",
"voiceId": "YOUR_MINIMAX_VOICE_ID",
"modelId": "minimax-2.8-turbo",
"emotion": "happy"
}| Value | Meaning |
|---|---|
happy | Bright and positive |
sad | Subdued and sorrowful |
angry | Forceful and intense |
surprised | Unexpected, with a rising delivery |
fearful | Nervous or worried |
calm | Steady and relaxed |
disgusted | Displeased or rejecting |
Global emotion currently applies only to minimax-2.8-turbo, minimax-2.8-hd, minimax-2.6-turbo, and minimax-2.6-hd. Fish Audio and Qwen models do not apply this request field.
Omit emotion to let the model choose a natural delivery from the text and voice. Use the lowercase values exactly as shown; other values may cause the request to fail. The field applies to the whole request, and its strength can vary by voice.
For local expression changes with supported Fish Audio models, use fine-grained emotion. You can also use GET /api/openapi.json to view the latest supported values.
See TTS models, synchronous TTS, asynchronous TTS, and fine-grained emotion.