> ## 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. # Saaras v3 > "Stream real-time speech-to-text transcripts from Sarvam AI Saaras v3 over WebSocket with voice activity detection across 23 Indian languages. Session configuration is provided via query parameters on the WebSocket upgrade URL: `language-code`, `mode`, `sample_rate`, `input_audio_codec`, `high_vad_sensitivity`, `vad_signals`." ## AsyncAPI ````yaml api-reference/asyncapi/stt-sarvam.asyncapi.yaml /v1/stt/sarvam/saaras:v3 id: /v1/stt/sarvam/saaras:v3 title: Saaras v3 description: >- Stream real-time speech-to-text transcripts from Sarvam AI Saaras v3 over WebSocket with voice activity detection across 23 Indian languages. Session configuration is provided via query parameters on the WebSocket upgrade URL: `language-code`, `mode`, `sample_rate`, `input_audio_codec`, `high_vad_sensitivity`, `vad_signals`. servers: - id: production protocol: wss host: api.slng.ai bindings: [] variables: [] - id: staging protocol: wss host: stageapi.slng.ai bindings: [] variables: [] address: /v1/stt/sarvam/saaras:v3 parameters: [] bindings: - protocol: ws version: latest value: method: GET headers: type: object properties: X-World-Part-Override: type: string description: >- Target world part override. Auto-selected if not provided. Available world parts: `ap`. enum: - ap x-parser-schema-id: x-parser-schema-id: WebsocketHeadersSarvamSaarasV3 schemaProperties: - name: method type: string description: GET required: false - name: headers type: object required: false properties: - name: X-World-Part-Override type: string description: >- Target world part override. Auto-selected if not provided. Available world parts: `ap`. enumValues: - ap required: false operations: - &ref_1 id: sttSarvamSaarasV3ReceiveAudio title: Stt sarvam saaras v3 receive audio description: Send audio frame to Saaras v3 type: receive messages: - &ref_6 id: SarvamSaarasV3AudioFrame contentType: application/json payload: - name: Audio Frame description: >- Stream a base64-encoded audio chunk to Saaras v3 as a JSON message. type: object properties: - name: audio type: object required: true properties: - name: data type: string description: >- Base64-encoded audio bytes for this chunk. At 16 kHz linear16, a ~256 ms chunk is roughly 8192 bytes before base64. required: true - name: sample_rate type: integer description: Sample rate of the audio in Hz. required: true - name: encoding type: string description: >- Encoding of the audio bytes. Raw PCM variants are normalized to `audio/wav` before being forwarded to Sarvam. enumValues: - wav - linear16 - LINEAR16 - pcm_s16le - pcm_l16 - pcm_raw - audio/wav required: true headers: [] jsonPayloadSchema: type: object description: >- Audio frame envelope expected by Saaras v3. Send a JSON message whose `audio.data` is a base64-encoded PCM/WAV chunk together with the sample rate and encoding. The gateway forwards the frame to Sarvam after normalizing `encoding` values to `audio/wav` when the source is raw PCM. required: - audio properties: audio: type: object required: - data - sample_rate - encoding properties: data: type: string format: byte description: >- Base64-encoded audio bytes for this chunk. At 16 kHz linear16, a ~256 ms chunk is roughly 8192 bytes before base64. x-parser-schema-id: sample_rate: type: integer description: Sample rate of the audio in Hz. default: 16000 x-parser-schema-id: encoding: type: string description: >- Encoding of the audio bytes. Raw PCM variants are normalized to `audio/wav` before being forwarded to Sarvam. default: wav enum: - wav - linear16 - LINEAR16 - pcm_s16le - pcm_l16 - pcm_raw - audio/wav x-parser-schema-id: x-parser-schema-id: x-parser-schema-id: title: Audio Frame description: Stream a base64-encoded audio chunk to Saaras v3 as a JSON message. example: |- { "audio": { "data": "UklGRiQAAABXQVZFZm10IBAAAAABAAEAQB8AAEAfAAABAAgAZGF0YQAAAAA=", "sample_rate": 16000, "encoding": "wav" } } bindings: [] extensions: - id: x-parser-unique-object-id value: SarvamSaarasV3AudioFrame bindings: [] extensions: &ref_0 - id: x-parser-unique-object-id value: /v1/stt/sarvam/saaras:v3 - &ref_2 id: sttSarvamSaarasV3ReceiveFlush title: Stt sarvam saaras v3 receive flush description: Force-finalize buffered audio on Saaras v3 type: receive messages: - &ref_7 id: SarvamSaarasV3FlushSignal contentType: application/json payload: - name: Flush Signal description: Force-finalize buffered audio without closing the connection. type: object properties: - name: type type: string description: flush required: true headers: [] jsonPayloadSchema: type: object description: >- Mid-stream flush. The gateway translates this to Saaras's `finalize` control message so the server emits a final transcript for the audio buffered so far. The WebSocket stays open for further audio. required: - type properties: type: type: string const: flush x-parser-schema-id: x-parser-schema-id: title: Flush Signal description: Force-finalize buffered audio without closing the connection. example: |- { "type": "flush" } bindings: [] extensions: - id: x-parser-unique-object-id value: SarvamSaarasV3FlushSignal bindings: [] extensions: *ref_0 - &ref_3 id: sttSarvamSaarasV3SendData title: Stt sarvam saaras v3 send data description: Receive transcript from Saaras v3 type: send messages: - &ref_8 id: SarvamSaarasV3TranscriptMessage contentType: application/json payload: - name: Transcript description: Saaras v3 final transcript chunk for the most recent utterance. type: object properties: - name: type type: string description: data required: true - name: data type: object required: true properties: - name: transcript type: string description: >- Transcribed text for the utterance, in the language determined by the `language-code` query parameter. required: true headers: [] jsonPayloadSchema: type: object description: >- Final transcript event emitted by Saaras v3 after an utterance ends or after a flush. Saaras v3 emits final transcripts only — there are no partial/interim results on this channel. required: - type - data properties: type: type: string const: data x-parser-schema-id: data: type: object required: - transcript properties: transcript: type: string description: >- Transcribed text for the utterance, in the language determined by the `language-code` query parameter. x-parser-schema-id: x-parser-schema-id: x-parser-schema-id: title: Transcript description: Saaras v3 final transcript chunk for the most recent utterance. example: |- { "type": "data", "data": { "transcript": "నమస్తే ఇది తెలుగు పరీక్ష" } } bindings: [] extensions: - id: x-parser-unique-object-id value: SarvamSaarasV3TranscriptMessage bindings: [] extensions: *ref_0 - &ref_4 id: sttSarvamSaarasV3SendEvents title: Stt sarvam saaras v3 send events description: Receive VAD events from Saaras v3 type: send messages: - &ref_9 id: SarvamSaarasV3EventsMessage contentType: application/json payload: - name: VAD Event description: Voice activity detection signal from Saaras v3. type: object properties: - name: type type: string description: events required: true - name: data type: object required: true properties: - name: signal_type type: string description: Which VAD boundary this event marks. enumValues: - START_SPEECH - END_SPEECH required: true - name: occured_at type: string description: ISO-8601 timestamp at which Saaras observed the boundary. required: false headers: [] jsonPayloadSchema: type: object description: >- Voice activity event emitted when `vad_signals=true` is set on the upgrade query string. `START_SPEECH` marks the start of an utterance and `END_SPEECH` marks the end. required: - type - data properties: type: type: string const: events x-parser-schema-id: data: type: object required: - signal_type properties: signal_type: type: string description: Which VAD boundary this event marks. enum: - START_SPEECH - END_SPEECH x-parser-schema-id: occured_at: type: string format: date-time description: ISO-8601 timestamp at which Saaras observed the boundary. x-parser-schema-id: x-parser-schema-id: x-parser-schema-id: title: VAD Event description: Voice activity detection signal from Saaras v3. example: |- { "type": "events", "data": { "signal_type": "END_SPEECH", "occured_at": "2026-05-18T12:50:56.000Z" } } bindings: [] extensions: - id: x-parser-unique-object-id value: SarvamSaarasV3EventsMessage bindings: [] extensions: *ref_0 - &ref_5 id: sttSarvamSaarasV3SendError title: Stt sarvam saaras v3 send error description: Receive error from Saaras v3 type: send messages: - &ref_10 id: SarvamSaarasV3ErrorMessage contentType: application/json payload: - name: Error description: Error from Saaras v3 or the SLNG gateway. type: object properties: - name: type type: string description: error required: true - name: data type: object required: true properties: - name: message type: string description: Human-readable error description. required: true - name: code type: string description: >- Optional machine-readable error code when the failure originates from the gateway. required: false headers: [] jsonPayloadSchema: type: object description: >- Error message from the upstream Sarvam session or from the SLNG gateway when a request is malformed, an authentication or quota check fails, or the upstream socket disconnects unexpectedly. required: - type - data properties: type: type: string const: error x-parser-schema-id: data: type: object required: - message properties: message: type: string description: Human-readable error description. x-parser-schema-id: code: type: string description: >- Optional machine-readable error code when the failure originates from the gateway. x-parser-schema-id: x-parser-schema-id: x-parser-schema-id: title: Error description: Error from Saaras v3 or the SLNG gateway. example: |- { "type": "error", "data": { "message": "Invalid audio payload" } } bindings: [] extensions: - id: x-parser-unique-object-id value: SarvamSaarasV3ErrorMessage bindings: [] extensions: *ref_0 sendOperations: - *ref_1 - *ref_2 receiveOperations: - *ref_3 - *ref_4 - *ref_5 sendMessages: - *ref_6 - *ref_7 receiveMessages: - *ref_8 - *ref_9 - *ref_10 extensions: - id: x-parser-unique-object-id value: /v1/stt/sarvam/saaras:v3 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: [] ````