Fish TTS 2.1 Pro
Synthesize speech with SLNG-hosted Fish TTS 2.1 Pro in a single HTTP request. The response is binary audio in the format set by format (MP3 by default).
Send application/json to synthesize with a saved voice via reference_id, or application/msgpack to clone a voice from inline reference audio supplied in references. For incremental, low-latency synthesis, connect to the WebSocket channel at the same path instead.
Authorizations
API key issued by SLNG. Pass as Authorization: Bearer <token>.
Headers
Target world part override. Auto-selected if not provided.
au, eu Body
Request body for SLNG-hosted Fish TTS 2.1 Pro synthesis. Shared by the application/json and application/msgpack content types; only MessagePack can carry inline reference audio in references.
Text to synthesize. S2-family variants support multi-speaker dialogue markers.
1Fish Audio voice-model ID. For multi-speaker synthesis, provide one ID per speaker as an array.
Inline reference audio for voice cloning. MessagePack-only, as the entries carry binary audio.
Prosody controls.
Sampling temperature.
0 <= x <= 1Nucleus sampling probability mass.
0 <= x <= 1Target chunk length for synthesis.
100 <= x <= 300Whether to normalize input text before synthesis.
Output audio format.
wav, pcm, mp3, opus Output sample rate in Hz. Defaults to the provider default for the chosen format when omitted.
8000, 16000, 24000, 32000, 44100, 48000 MP3 bitrate in kbps (used when format is mp3).
64, 128, 192 Opus bitrate in bps (used when format is opus). -1000 selects the provider default.
-1000, 24000, 32000, 48000, 64000 Latency/quality trade-off.
low, normal, balanced Maximum number of new tokens to generate.
Penalty applied to repeated tokens.
Minimum chunk length for synthesis.
0 <= x <= 100Whether generation is conditioned on previously generated chunks.
Early-stop threshold for generation.
0 <= x <= 1Optional feature flags.
Response
Synthesis successful.
Binary audio data.