Fish Audio · Speech Audio

FishAudio OpenAPI / Text to Speech

Turn text into speech with Fish Audio (models s2.1-pro, s2-pro and drama-3). Returns an audio file. Up to 5,000 characters per call, about 5 minutes of speech. Pass reference_id (a voice from fish-audio-model-get) to pick a voice. With model drama-3 you can direct the voice in plain words inside square brackets, for example [whisper] or [laughs], and write multi-speaker dialogue. Charged per character of text, nothing for a failed call.

Live data

You are looking at the live public catalog.

unverifiedRequest shape unavailableUnverified

Published API evidence

Public metadata only. Response bodies, credentials, and internal review notes are never displayed.

No verification date is claimed. No response capture is claimed.

Request parameters

chunk_length body · optional
Text segment size for processing.
condition_on_previous_chunks body · optional
Use previous audio as context for voice consistency.
early_stop_threshold body · optional
Early stopping threshold for batch processing.
features body · optional
Optional request-scoped TTS feature flags forwarded verbatim to the inference backend. Use ["quality-guard"] to enable the quality guard for this synthesis request. Feature availability is determined by the inference backend.
format body · optional
Audio format of the file you get back.
language body · optional
latency body · optional
Latency-quality trade-off. normal: best quality, balanced: reduced latency, low: lowest latency.
max_new_tokens body · optional
Maximum audio tokens to generate per text chunk.
min_chunk_length body · optional
Minimum characters before splitting into a new chunk.
mp3_bitrate body · optional
MP3 bitrate in kbps. Only applies when format is mp3.
normalize body · optional
Rewrites numbers, dates, units and symbols as spoken words before synthesis, improving stability. Uses `language` when it is set and otherwise detects the language of each text chunk.
opus_bitrate body · optional
Opus bitrate in bps. -1000 for automatic. Only applies when format is opus.
pronunciation_dictionary body · optional
Reference form: list of pointers to platform pronunciation dictionaries. Each entry identifies a dictionary by `id` and `version`; the server resolves the contents from the platform service before forwarding them to the inference engine.
prosody body · optional
Controls for adjusting the prosody (rhythm and intonation) of generated speech.
reference_id body · optional
Single speaker: voice model ID string
references body · optional
Single speaker: array of reference audio samples
repetition_penalty body · optional
Penalty for repeating audio patterns. Values above 1.0 reduce repetition.
sample_rate body · optional
temperature body · optional
Controls expressiveness. Higher is more varied, lower is more consistent.
text body · required
Text to convert to speech, 1 to 5,000 characters.
top_p body · optional
Controls diversity via nucleus sampling.
model header · optional
Speech model. All three bill the same per character.

No allowlisted official documentation or pricing reference is available.