Fish Audio Docs
API ReferenceText to Speech

TTS Models

Review the supported modelId values, model characteristics, and selection guidance for text to speech.

TTS Models

Use modelId to select the TTS engine and version, and use voiceId to select the speaker. Send both fields in sync and async TTS requests so the generation and billing model are explicit.

Supported modelId values

modelIdBrandVersionBest suited for
fishaudio-s21proFish AudioS2.1 ProHigher-quality production generation
fishaudio-s21pro-flashFish AudioS2.1 Pro FlashDefault; faster with lower credit consumption
fishaudio-s2proFish AudioS2 ProWorkflows that target S2 Pro
fishaudio-s1Fish AudioS1Existing integrations that still depend on S1
minimax-2.8-turboMiniMaxSpeech 2.8 TurboDefault MiniMax option focused on speed
minimax-2.8-hdMiniMaxSpeech 2.8 HDHigher-quality MiniMax output
minimax-2.6-turboMiniMaxSpeech 2.6 TurboCompatibility with 2.6 Turbo workflows
minimax-2.6-hdMiniMaxSpeech 2.6 HDCompatibility with 2.6 HD workflows
qwen3-tts-flashQwenTTS 3 FlashFast generation with Qwen voices

Treat the TTS schema returned by GET /api/openapi.json as the source of truth. It reflects the current public contract when model availability, account access, or brand capabilities change.

Choosing a model

  • Choose a modelId from the same brand as the selected voiceId. For example, use a fishaudio-* model with a Fish Audio voice.
  • Fish Audio defaults to fishaudio-s21pro-flash when modelId is omitted.
  • MiniMax defaults to minimax-2.8-turbo, and Qwen defaults to qwen3-tts-flash.
  • Models can differ in quality, latency, expression, and credit consumption. Test a short sample before starting a large batch.

Request example

Sync and async TTS requests both use modelId:

{
  "text": "Hello, this is a TTS model test.",
  "voiceId": "00a1b221-6137-4b73-ad62-b0cbce134167",
  "modelId": "fishaudio-s21pro-flash",
  "format": "mp3",
  "speed": 1
}

Endpoint mapping

EndpointModel field
POST /api/open/v1/speech/ttsmodelId
POST /api/open/v1/speech/tts/jobsmodelId
OpenAI-compatible POST /v1/audio/speechmodel

Continue with Sync HTTP, Async Jobs, or OpenAI-Compatible TTS for complete request details.