> ## Documentation Index > Fetch the complete documentation index at: https://docs.slng.ai/llms.txt > Use this file to discover all available pages before exploring further. # Unified TTS > Stream synthesized audio from any WebSocket-capable TTS model over SLNG's native protocol. Send `init` (the model comes from the route path; `model` in the body is advisory), `text` frames, and `flush` to end a turn. The gateway returns `ready`, raw binary audio frames in the init encoding and sample rate, a bare `audio_end` after each flushed turn, `cleared` after `clear`, and flat `error` frames. There is no JSON `audio_chunk`, no `flushed` event, and no `data` envelope on any message. Encodings and sample rates are model-specific: Deepgram Aura accepts linear16 (8000–48000 Hz), mulaw, and alaw (8000/16000 Hz) over WebSocket only; Fish Audio accepts linear16/pcm/wav (8000–44100 Hz), mp3 (32000/44100 Hz), and opus (48000 Hz). ## AsyncAPI ````yaml api-reference/asyncapi/bridges-unmute.asyncapi.yaml /v1/bridges/unmute/tts/{model_variant} id: /v1/bridges/unmute/tts/{model_variant} title: Unmute TTS Bridge description: >- Stream synthesized audio from any WebSocket-capable TTS model over SLNG's native protocol. Send `init` (the model comes from the route path; `model` in the body is advisory), `text` frames, and `flush` to end a turn. The gateway returns `ready`, raw binary audio frames in the init encoding and sample rate, a bare `audio_end` after each flushed turn, `cleared` after `clear`, and flat `error` frames. There is no JSON `audio_chunk`, no `flushed` event, and no `data` envelope on any message. Encodings and sample rates are model-specific: Deepgram Aura accepts linear16 (8000–48000 Hz), mulaw, and alaw (8000/16000 Hz) over WebSocket only; Fish Audio accepts linear16/pcm/wav (8000–44100 Hz), mp3 (32000/44100 Hz), and opus (48000 Hz). servers: - id: production protocol: wss host: api.slng.ai bindings: [] variables: [] address: /v1/bridges/unmute/tts/{model_variant} parameters: [] bindings: - protocol: ws version: latest value: method: GET schemaProperties: - name: method type: string description: GET required: false operations: - &ref_2 id: unmuteTtsBridgeReceiveInit title: Unmute tts bridge receive init description: Initialize session on Unmute TTS Bridge type: receive messages: - &ref_14 id: UnmuteTtsInitMessage contentType: application/json payload: - name: Init (Unmute TTS) description: Initialize Unmute TTS session with voice and configuration. type: object properties: - name: type type: string description: init required: true - name: model type: string description: >- Advisory and optional. Never honoured or forwarded; the route path selects the model. required: false - name: voice type: string description: Voice identifier for the selected model. required: false - name: pronunciation type: object description: >- Optional pronunciation rewrite. `mode` is always `rewrite`; supply exactly one of `name` (a dictionary name, 1–128 characters) or `dictionary_id`. required: false properties: - name: mode type: string description: rewrite required: true - name: name type: string required: false - name: dictionary_id type: string required: false - name: config type: object description: Synthesis configuration. required: false properties: - name: sample_rate type: integer description: Audio sample rate in Hz. enumValues: - 8000 - 16000 - 22050 - 24000 - 32000 - 44100 - 48000 required: false - name: encoding type: string description: >- Union across bridge targets; each route accepts only its own subset. Deepgram Aura over WebSocket accepts linear16, mulaw, and alaw only (mp3/opus/flac are HTTP-only for Aura). Fish Audio accepts linear16 (alias of pcm), pcm, wav, mp3, and opus. enumValues: - linear16 - mp3 - opus - mulaw - alaw - flac - aac - wav - pcm required: false - name: language type: string description: ISO-639-1 language code. enumValues: - en - ja - zh - de - hi - fr - ko - pt - it - es - id - nl - tr - fil - pl - sv - bg - ro - ar - cs - el - fi - hr - ms - sk - da - ta - uk - ru - hu - 'no' - vi required: false - name: speed type: number description: >- Optional speaking-speed hint in the selected model's units; models without a speed control accept and ignore it. required: false - name: pronunciation type: object description: >- Optional pronunciation rewrite. `mode` is always `rewrite`; supply exactly one of `name` (a dictionary name, 1–128 characters) or `dictionary_id`. required: false properties: - name: mode type: string description: rewrite required: true - name: name type: string required: false - name: dictionary_id type: string required: false headers: [] jsonPayloadSchema: type: object description: >- Opens the provider session. The model comes from the route path; `model` in the body is advisory and discarded. `config.encoding` and `config.sample_rate` are validated against the selected model before any provider connection and rejected with a `config_error` otherwise. required: - type properties: type: type: string const: init x-parser-schema-id: model: type: string description: >- Advisory and optional. Never honoured or forwarded; the route path selects the model. x-parser-schema-id: voice: type: string description: Voice identifier for the selected model. x-parser-schema-id: pronunciation: type: object description: >- Optional pronunciation rewrite. `mode` is always `rewrite`; supply exactly one of `name` (a dictionary name, 1–128 characters) or `dictionary_id`. required: - mode properties: mode: type: string const: rewrite x-parser-schema-id: name: type: string minLength: 1 maxLength: 128 x-parser-schema-id: dictionary_id: type: string minLength: 1 x-parser-schema-id: x-parser-schema-id: config: type: object description: Synthesis configuration. properties: sample_rate: type: integer description: Audio sample rate in Hz. enum: - 8000 - 16000 - 22050 - 24000 - 32000 - 44100 - 48000 x-parser-schema-id: SampleRate encoding: type: string description: >- Union across bridge targets; each route accepts only its own subset. Deepgram Aura over WebSocket accepts linear16, mulaw, and alaw only (mp3/opus/flac are HTTP-only for Aura). Fish Audio accepts linear16 (alias of pcm), pcm, wav, mp3, and opus. enum: - linear16 - mp3 - opus - mulaw - alaw - flac - aac - wav - pcm x-parser-schema-id: language: type: string description: ISO-639-1 language code. enum: - en - ja - zh - de - hi - fr - ko - pt - it - es - id - nl - tr - fil - pl - sv - bg - ro - ar - cs - el - fi - hr - ms - sk - da - ta - uk - ru - hu - 'no' - vi x-parser-schema-id: LanguageCode speed: type: number description: >- Optional speaking-speed hint in the selected model's units; models without a speed control accept and ignore it. x-parser-schema-id: pronunciation: type: object description: >- Optional pronunciation rewrite. `mode` is always `rewrite`; supply exactly one of `name` (a dictionary name, 1–128 characters) or `dictionary_id`. required: - mode properties: mode: type: string const: rewrite x-parser-schema-id: name: type: string minLength: 1 maxLength: 128 x-parser-schema-id: dictionary_id: type: string minLength: 1 x-parser-schema-id: x-parser-schema-id: x-parser-schema-id: x-parser-schema-id: title: Init (Unmute TTS) description: Initialize Unmute TTS session with voice and configuration. example: |- { "type": "init", "voice": "aura-2-thalia-en", "config": { "sample_rate": 24000, "encoding": "linear16" } } bindings: [] extensions: - id: x-parser-unique-object-id value: UnmuteTtsInitMessage bindings: [] extensions: &ref_0 - id: x-parser-unique-object-id value: /v1/bridges/unmute/tts/{model_variant} - &ref_3 id: unmuteTtsBridgeReceiveText title: Unmute tts bridge receive text description: Send text to Unmute TTS Bridge type: receive messages: - &ref_15 id: UnmuteTtsTextMessage contentType: application/json payload: - name: Text (Unmute TTS) description: Send text to synthesize. type: object properties: - name: type type: string description: text required: true - name: text type: string description: Text to synthesize. required: true - name: flush type: boolean description: Flush the turn immediately after this text. required: false - name: partial type: boolean description: >- Marks this frame as the start or middle of a sentence that the next text frames continue. The gateway either sends it on or holds it and sends it joined to the next frame, whichever suits the provider. Held text is joined exactly as sent, with no added spaces. Cannot be combined with `flush: true`. A frame with `partial: true` may contain only `type`, `text`, `partial`, `pronunciation`, and `flush: false`. End the sentence with an ordinary text frame or a `flush`; if the connection closes first, held text is discarded. Where the gateway sends the fragment on, the provider receives it unchanged, so send at least a clause at a time: single words can be voiced one at a time. Send `partial` only when `ready` includes `accepts_partial_text: true`. required: false - name: pronunciation type: object description: >- Optional pronunciation rewrite. `mode` is always `rewrite`; supply exactly one of `name` (a dictionary name, 1–128 characters) or `dictionary_id`. required: false properties: - name: mode type: string description: rewrite required: true - name: name type: string required: false - name: dictionary_id type: string required: false headers: [] jsonPayloadSchema: type: object description: >- Text for the current turn. `flush: true` ends the turn with this text. Set `partial: true` on a frame that ends mid-sentence when more of the same sentence follows. required: - type - text properties: type: type: string const: text x-parser-schema-id: text: type: string description: Text to synthesize. x-parser-schema-id: flush: type: boolean description: Flush the turn immediately after this text. x-parser-schema-id: partial: type: boolean description: >- Marks this frame as the start or middle of a sentence that the next text frames continue. The gateway either sends it on or holds it and sends it joined to the next frame, whichever suits the provider. Held text is joined exactly as sent, with no added spaces. Cannot be combined with `flush: true`. A frame with `partial: true` may contain only `type`, `text`, `partial`, `pronunciation`, and `flush: false`. End the sentence with an ordinary text frame or a `flush`; if the connection closes first, held text is discarded. Where the gateway sends the fragment on, the provider receives it unchanged, so send at least a clause at a time: single words can be voiced one at a time. Send `partial` only when `ready` includes `accepts_partial_text: true`. x-parser-schema-id: pronunciation: type: object description: >- Optional pronunciation rewrite. `mode` is always `rewrite`; supply exactly one of `name` (a dictionary name, 1–128 characters) or `dictionary_id`. required: - mode properties: mode: type: string const: rewrite x-parser-schema-id: name: type: string minLength: 1 maxLength: 128 x-parser-schema-id: dictionary_id: type: string minLength: 1 x-parser-schema-id: x-parser-schema-id: x-parser-schema-id: title: Text (Unmute TTS) description: Send text to synthesize. example: |- { "type": "text", "text": "Hello, how can I help you?" } bindings: [] extensions: - id: x-parser-unique-object-id value: UnmuteTtsTextMessage bindings: [] extensions: *ref_0 - &ref_4 id: unmuteTtsBridgeReceiveFlush title: Unmute tts bridge receive flush description: Flush the current turn on Unmute TTS Bridge type: receive messages: - &ref_16 id: UnmuteTtsFlushMessage contentType: application/json payload: - name: Flush (Unmute TTS) description: End the current turn and synthesize buffered text. type: object properties: - name: type type: string description: flush required: true headers: [] jsonPayloadSchema: type: object description: >- Ends the current turn. Binary audio frames follow and the turn terminates with a bare `audio_end`; there is no `flushed` acknowledgement. required: - type properties: type: type: string const: flush x-parser-schema-id: x-parser-schema-id: title: Flush (Unmute TTS) description: End the current turn and synthesize buffered text. example: |- { "type": "flush" } bindings: [] extensions: - id: x-parser-unique-object-id value: UnmuteTtsFlushMessage bindings: [] extensions: *ref_0 - &ref_5 id: unmuteTtsBridgeReceiveClear title: Unmute tts bridge receive clear description: Discard queued output on Unmute TTS Bridge type: receive messages: - &ref_17 id: UnmuteTtsClearMessage contentType: application/json payload: - name: Clear (Unmute TTS) description: Discard queued output for the current turn. type: object properties: - name: type type: string description: clear required: true headers: [] jsonPayloadSchema: type: object description: >- Cancels the active provider turn and discards queued output. The gateway replies with `cleared`. required: - type properties: type: type: string const: clear x-parser-schema-id: x-parser-schema-id: title: Clear (Unmute TTS) description: Discard queued output for the current turn. example: |- { "type": "clear" } bindings: [] extensions: - id: x-parser-unique-object-id value: UnmuteTtsClearMessage bindings: [] extensions: *ref_0 - &ref_6 id: unmuteTtsBridgeReceiveCancel title: Unmute tts bridge receive cancel description: Cancel the current synthesis turn on Unmute TTS Bridge type: receive messages: - &ref_18 id: UnmuteTtsCancelMessage contentType: application/json payload: - name: Cancel (Unmute TTS) description: Cancel the current synthesis turn. type: object properties: - name: type type: string description: cancel required: true headers: [] jsonPayloadSchema: type: object description: >- Cancels the current synthesis turn without a dedicated acknowledgement; the session stays open for the next turn. required: - type properties: type: type: string const: cancel x-parser-schema-id: x-parser-schema-id: title: Cancel (Unmute TTS) description: Cancel the current synthesis turn. example: |- { "type": "cancel" } bindings: [] extensions: - id: x-parser-unique-object-id value: UnmuteTtsCancelMessage bindings: [] extensions: *ref_0 - &ref_7 id: unmuteTtsBridgeReceiveClose title: Unmute tts bridge receive close description: Close session on Unmute TTS Bridge type: receive messages: - &ref_19 id: UnmuteTtsCloseMessage contentType: application/json payload: - name: Close (Unmute TTS) description: Close the session. type: object properties: - name: type type: string description: close required: true headers: [] jsonPayloadSchema: type: object description: Ends the session after in-flight audio is delivered. required: - type properties: type: type: string const: close x-parser-schema-id: x-parser-schema-id: title: Close (Unmute TTS) description: Close the session. example: |- { "type": "close" } bindings: [] extensions: - id: x-parser-unique-object-id value: UnmuteTtsCloseMessage bindings: [] extensions: *ref_0 - &ref_8 id: unmuteTtsBridgeReceiveKeepalive title: Unmute tts bridge receive keepalive description: Send keepalive on Unmute TTS Bridge type: receive messages: - &ref_20 id: UnmuteTtsKeepaliveMessage contentType: application/json payload: - name: Keepalive (Unmute TTS) description: Keep an idle session open. type: object properties: - name: type type: string description: keepalive required: true headers: [] jsonPayloadSchema: type: object description: >- Resets the gateway idle timeout and is forwarded upstream only for providers with a native keepalive frame (Sarvam). It produces no reply and never extends the maximum session duration. required: - type properties: type: type: string const: keepalive x-parser-schema-id: x-parser-schema-id: title: Keepalive (Unmute TTS) description: Keep an idle session open. example: |- { "type": "keepalive" } bindings: [] extensions: - id: x-parser-unique-object-id value: UnmuteTtsKeepaliveMessage bindings: [] extensions: *ref_0 - &ref_9 id: unmuteTtsBridgeSendReady title: Unmute tts bridge send ready description: Receive ready signal from Unmute TTS Bridge type: send messages: - &ref_21 id: UnmuteTtsReadyMessage contentType: application/json payload: - name: Ready (Unmute TTS) description: Unmute TTS session is ready. type: object properties: - name: type type: string description: ready required: true - name: session_id type: string description: Gateway request identifier for this session. required: true - name: slng_request_id type: string description: >- SLNG request identifier (same value as `session_id`) for log correlation. required: true - name: accepts_partial_text type: boolean description: >- `true` when this gateway honours `partial` on text frames. Absent when it does not, in which case do not send `partial`. required: false headers: [] jsonPayloadSchema: type: object description: >- Sent once after the upstream session is open. `session_id` and `slng_request_id` carry the same gateway request identifier. required: - type - session_id - slng_request_id properties: type: type: string const: ready x-parser-schema-id: session_id: type: string description: Gateway request identifier for this session. x-parser-schema-id: slng_request_id: type: string description: >- SLNG request identifier (same value as `session_id`) for log correlation. x-parser-schema-id: accepts_partial_text: type: boolean description: >- `true` when this gateway honours `partial` on text frames. Absent when it does not, in which case do not send `partial`. x-parser-schema-id: x-parser-schema-id: title: Ready (Unmute TTS) description: Unmute TTS session is ready. example: |- { "type": "ready", "session_id": "1e1a9598-077e-4d9d-aba6-78af6363f27c", "slng_request_id": "1e1a9598-077e-4d9d-aba6-78af6363f27c", "accepts_partial_text": true } bindings: [] extensions: - id: x-parser-unique-object-id value: UnmuteTtsReadyMessage bindings: [] extensions: *ref_0 - &ref_10 id: unmuteTtsBridgeSendBinaryAudio title: Unmute tts bridge send binary audio description: Receive a raw binary audio frame from Unmute TTS Bridge type: send messages: - &ref_22 id: UnmuteTtsBinaryAudio contentType: application/octet-stream payload: - name: Binary Audio Frame description: Raw binary audio frame from Unmute TTS. type: object properties: - name: type type: string description: >- Placeholder discriminator. The actual message is a raw binary WebSocket frame, not JSON. required: true headers: [] jsonPayloadSchema: type: object description: >- Synthesized audio is delivered only as raw binary WebSocket frames in the encoding and sample rate selected at init. This is NOT a JSON message and there is no base64 `audio_chunk` envelope. The object schema below exists only because the AsyncAPI tooling requires one. required: - type properties: type: type: string const: binary description: >- Placeholder discriminator. The actual message is a raw binary WebSocket frame, not JSON. x-parser-schema-id: x-parser-schema-id: title: Binary Audio Frame description: Raw binary audio frame from Unmute TTS. example: >- "0000FF00000000FF00000000010101010101010100000000FFFFFFFFFFFEFEFDFEFEFEFEFDFDFEFEFEFEFEFEFEFEFEFFFFFFFFFEFEFEFEFF0001000001020303030303030303030201010000FFFFFEFDFDFDFDFEFFFFFFFF0001020303020201000000FFFDFCFBFAFAFBFAF9F8F7F7F7F6F6F4F2F2F3F7FC000406090F14191A19181715110E0A05FEF9F6F3F0EEECEBEBECEEF2F6F9FC0005090D0F101010100E0C080401" bindings: [] extensions: - id: x-parser-unique-object-id value: UnmuteTtsBinaryAudio bindings: [] extensions: *ref_0 - &ref_11 id: unmuteTtsBridgeSendCleared title: Unmute tts bridge send cleared description: Receive clear acknowledgement from Unmute TTS Bridge type: send messages: - &ref_23 id: UnmuteTtsClearedMessage contentType: application/json payload: - name: Cleared (Unmute TTS) description: Queued output was discarded. type: object properties: - name: type type: string description: cleared required: true headers: [] jsonPayloadSchema: type: object description: >- Acknowledges a client `clear`; queued provider output for the turn was discarded. required: - type properties: type: type: string const: cleared x-parser-schema-id: x-parser-schema-id: title: Cleared (Unmute TTS) description: Queued output was discarded. example: |- { "type": "cleared" } bindings: [] extensions: - id: x-parser-unique-object-id value: UnmuteTtsClearedMessage bindings: [] extensions: *ref_0 - &ref_12 id: unmuteTtsBridgeSendAudioEnd title: Unmute tts bridge send audio end description: Receive bare audio_end turn boundary from Unmute TTS Bridge type: send messages: - &ref_24 id: UnmuteTtsAudioEndMessage contentType: application/json payload: - name: Audio End (Unmute TTS) description: Bare turn boundary after the last audio frame. type: object properties: - name: type type: string description: audio_end required: true headers: [] jsonPayloadSchema: type: object description: >- Sent after the last binary audio frame of a flushed turn, on cache replay, and when the provider transport closes cleanly. The frame is exactly `{"type":"audio_end"}`: no `duration`, no payload, no nested `data`. required: - type properties: type: type: string const: audio_end x-parser-schema-id: x-parser-schema-id: title: Audio End (Unmute TTS) description: Bare turn boundary after the last audio frame. example: |- { "type": "audio_end" } bindings: [] extensions: - id: x-parser-unique-object-id value: UnmuteTtsAudioEndMessage bindings: [] extensions: *ref_0 - &ref_13 id: unmuteTtsBridgeSendError title: Unmute tts bridge send error description: Receive error from Unmute TTS Bridge type: send messages: - &ref_25 id: UnmuteTtsErrorMessage contentType: application/json payload: - name: Error (Unmute TTS) description: Error occurred during Unmute TTS session. type: object properties: - name: type type: string description: error required: true - name: code type: &ref_1 - string - integer description: >- Error code. Gateway codes are strings; a provider-forwarded code keeps the provider's value, which may be numeric. required: true - name: message type: string description: Human-readable error description; wording may change. required: true - name: slng_request_id type: string description: SLNG request identifier, present only on pronunciation errors. required: false headers: [] jsonPayloadSchema: type: object description: >- Flat error frame; there is no nested `data` envelope. Provider errors are forwarded with the provider's own `code` value (which may be numeric) or `provider_error` when the provider gave none. Gateway-generated TTS session errors use string codes such as `config_error`, `provider_error`, `rate_limit_exceeded`, `payload_too_large`, `processing_error`, `watermarking_error`, `unsupported_watermark_format`, and `invalid_pronunciation`. `slng_request_id` is present only on pronunciation errors; use the `ready` message for correlation otherwise. required: - type - code - message properties: type: type: string const: error x-parser-schema-id: code: type: *ref_1 description: >- Error code. Gateway codes are strings; a provider-forwarded code keeps the provider's value, which may be numeric. x-parser-schema-id: message: type: string description: Human-readable error description; wording may change. x-parser-schema-id: slng_request_id: type: string description: SLNG request identifier, present only on pronunciation errors. x-parser-schema-id: x-parser-schema-id: title: Error (Unmute TTS) description: Error occurred during Unmute TTS session. example: |- { "type": "error", "code": "provider_error", "message": "Upstream service unavailable" } bindings: [] extensions: - id: x-parser-unique-object-id value: UnmuteTtsErrorMessage bindings: [] extensions: *ref_0 sendOperations: - *ref_2 - *ref_3 - *ref_4 - *ref_5 - *ref_6 - *ref_7 - *ref_8 receiveOperations: - *ref_9 - *ref_10 - *ref_11 - *ref_12 - *ref_13 sendMessages: - *ref_14 - *ref_15 - *ref_16 - *ref_17 - *ref_18 - *ref_19 - *ref_20 receiveMessages: - *ref_21 - *ref_22 - *ref_23 - *ref_24 - *ref_25 extensions: - id: x-parser-unique-object-id value: /v1/bridges/unmute/tts/{model_variant} securitySchemes: - id: bearer name: bearer type: http description: > API key issued by SLNG. Pass as `Authorization: Bearer ` in the WebSocket upgrade request headers. scheme: bearer extensions: [] ```` This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.