Synchronous Text to Speech v2

HTTP TTS v2 is maintained for existing clients. New integrations should use HTTP TTS v3, which provides a strict field contract, model capabilities, and recoverable asynchronous jobs.

Endpoint

POST /api/open/v2/speech/tts

Full URL:

https://fishaudio.org/api/open/v2/speech/tts

Authentication and request headers

Authorization: Bearer FISHAUDIO_API_KEY
Content-Type: application/json
X-Request-Id: YOUR_STABLE_REQUEST_ID

Keep the API key on a trusted server. Success is binary audio; errors are JSON.

Request

{
  "text": "Hello from Fish Audio.",
  "voiceId": "00a1b221-6137-4b73-ad62-b0cbce134167",
  "modelId": "fishaudio-s21pro-flash",
  "format": "mp3",
  "speed": 1,
  "volume": 0,
  "stability": 1,
  "similarity": 1
}

Field	Type	Required	Notes
`text`	string	Yes	1–10,000 characters
`voiceId`	string	Yes	Public or personal voice ID
`modelId`	string	No	Public engine model ID; omission uses voice/account defaults
`format`	string	No	`mp3`, `wav`, or `ogg`; default `mp3`
`speed`	number	No	`0.5`–`2`; default `1`
`volume`	number	No	`-20`–`20`; default `0`
`pitch`	number	No	`-12`–`12`
`stability`	number	No	`0.5`–`1.5`
`similarity`	number	No	`0.5`–`1.5`
`language`	string	No	Language hint
`emotion`	string	No	Global emotion ID
`instruction`	string	No	Qwen style instruction, up to 1,600 characters
`textNormalization`	boolean	No	Structured-text pronunciation normalization

V2 also accepts historical fields such as engineModelId, version, qwenModel, and minimaxEmotion. They exist only for old clients. Do not use them in new code or mix them into v3 requests. V2 has no provider-neutral capability catalog.

curl example

curl "https://fishaudio.org/api/open/v2/speech/tts" \
  -H "Authorization: Bearer $FISHAUDIO_API_KEY" \
  -H "Content-Type: application/json" \
  -H "X-Request-Id: legacy-tts-001" \
  -d '{"text":"Hello","voiceId":"00a1b221-6137-4b73-ad62-b0cbce134167","modelId":"fishaudio-s21pro-flash","format":"mp3"}' \
  --output speech.mp3

Successful response

HTTP/1.1 200 OK
Content-Type: audio/mpeg
X-Request-Id: legacy-tts-001
X-OpenAPI-Quota-Remaining: 99994
X-OpenAPI-Credits-Used: 6

<binary audio data>

Errors and retries

Errors use a flat JSON shape:

{
  "code": "ERR_REQUEST_ID_CONFLICT",
  "message": "Request id was reused with a different request",
  "requestId": "legacy-tts-001"
}

Status	Meaning
`400`	Invalid request or voice/model combination
`401`	Invalid API key
`402`	Insufficient API quota
`409`	Request ID active, completed, or reused with different input
`413`	HTTP request body exceeds the limit
`429`	Rate limited; honor `Retry-After`
`500`	Generation or settlement failure

Reuse X-Request-Id only for an identical retry. Lost synchronous audio cannot be replayed; workflows that require recovery should migrate to v3 Jobs.

Migrate to v3

Keep text, voiceId, and generic controls; remove historical provider fields; and explicitly send a modelId returned by capabilities. See the migration guide for the complete mapping.