Deepgram Text-to-Speech Streaming Events

asyncapi: 2.6.0
info:
  title: Deepgram Text-to-Speech Streaming Events
  description: >-
    The Deepgram Text-to-Speech streaming API provides real-time speech
    synthesis over a WebSocket connection. Text is sent as JSON messages
    and audio data is returned as binary WebSocket messages, enabling
    continuous streaming text-to-speech for conversational AI applications,
    voice agents, and real-time voice interfaces.
  version: '1.0'
  contact:
    name: Deepgram Support
    url: https://developers.deepgram.com
servers:
  production:
    url: 'wss://api.deepgram.com/v1/speak'
    protocol: wss
    description: >-
      Deepgram production WebSocket server for real-time text-to-speech
      streaming. Connect with query parameters to configure the voice model,
      encoding, and sample rate.
    security:
      - bearerAuth: []
  eu:
    url: 'wss://api.eu.deepgram.com/v1/speak'
    protocol: wss
    description: >-
      Deepgram EU WebSocket server for real-time text-to-speech streaming.
    security:
      - bearerAuth: []
channels:
  /v1/speak:
    description: >-
      WebSocket channel for real-time text-to-speech streaming. The client
      sends text as JSON messages and receives synthesized audio as binary
      frames. Connection parameters include model, encoding, sample_rate,
      and container settings.
    publish:
      operationId: sendTextForSpeech
      summary: Send text for speech synthesis
      description: >-
        Client sends JSON messages containing text to be synthesized into
        speech. Supports continuous streaming of text segments.
      message:
        oneOf:
          - $ref: '#/components/messages/TextInput'
          - $ref: '#/components/messages/Flush'
          - $ref: '#/components/messages/Reset'
          - $ref: '#/components/messages/Close'
    subscribe:
      operationId: receiveSpeechAudio
      summary: Receive synthesized speech audio
      description: >-
        Server sends binary audio frames and JSON control messages as
        speech is synthesized from the input text.
      message:
        oneOf:
          - $ref: '#/components/messages/AudioData'
          - $ref: '#/components/messages/Flushed'
          - $ref: '#/components/messages/Warning'
          - $ref: '#/components/messages/TTSError'
components:
  securitySchemes:
    bearerAuth:
      type: http
      scheme: bearer
      description: >-
        Deepgram API key passed as a token query parameter or Authorization
        header when establishing the WebSocket connection.
  messages:
    TextInput:
      name: TextInput
      title: Text Input
      summary: Text to synthesize into speech
      description: >-
        JSON message containing text to be converted to speech audio. Text
        is synthesized incrementally as it is received.
      contentType: application/json
      payload:
        $ref: '#/components/schemas/TextInputPayload'
    Flush:
      name: Flush
      title: Flush
      summary: Flush pending text
      description: >-
        Signals the server to immediately synthesize any buffered text and
        return audio for it.
      contentType: application/json
      payload:
        $ref: '#/components/schemas/FlushPayload'
    Reset:
      name: Reset
      title: Reset
      summary: Reset the synthesis state
      description: >-
        Resets the text-to-speech synthesis state, clearing any buffered
        text that has not yet been synthesized.
      contentType: application/json
      payload:
        $ref: '#/components/schemas/ResetPayload'
    Close:
      name: Close
      title: Close
      summary: Close the streaming session
      description: >-
        Signals the server to finalize synthesis and close the connection
        after all pending audio has been returned.
      contentType: application/json
      payload:
        $ref: '#/components/schemas/ClosePayload'
    AudioData:
      name: AudioData
      title: Audio Data
      summary: Synthesized speech audio data
      description: >-
        Binary WebSocket message containing synthesized speech audio in the
        encoding format configured at connection time.
      contentType: application/octet-stream
      payload:
        type: string
        format: binary
        description: >-
          Raw binary audio data in the configured encoding format.
    Flushed:
      name: Flushed
      title: Flushed
      summary: Flush confirmation
      description: >-
        Confirmation that all buffered text has been synthesized and the
        corresponding audio has been sent.
      contentType: application/json
      payload:
        $ref: '#/components/schemas/FlushedPayload'
    Warning:
      name: Warning
      title: Warning
      summary: Warning message
      description: >-
        Non-fatal warning about the streaming session.
      contentType: application/json
      payload:
        $ref: '#/components/schemas/WarningPayload'
    TTSError:
      name: TTSError
      title: Error
      summary: Error message
      description: >-
        Error event indicating an issue with the text-to-speech session.
      contentType: application/json
      payload:
        $ref: '#/components/schemas/TTSErrorPayload'
  schemas:
    TextInputPayload:
      type: object
      required:
        - type
        - text
      properties:
        type:
          type: string
          const: Speak
          description: >-
            Message type identifier.
        text:
          type: string
          description: >-
            Text content to synthesize into speech.
    FlushPayload:
      type: object
      required:
        - type
      properties:
        type:
          type: string
          const: Flush
          description: >-
            Message type identifier.
    ResetPayload:
      type: object
      required:
        - type
      properties:
        type:
          type: string
          const: Reset
          description: >-
            Message type identifier.
    ClosePayload:
      type: object
      required:
        - type
      properties:
        type:
          type: string
          const: Close
          description: >-
            Message type identifier.
    FlushedPayload:
      type: object
      properties:
        type:
          type: string
          const: Flushed
          description: >-
            Message type identifier.
        sequence_id:
          type: integer
          description: >-
            Sequence identifier for the flush operation.
    WarningPayload:
      type: object
      properties:
        type:
          type: string
          const: Warning
          description: >-
            Message type identifier.
        warn_code:
          type: string
          description: >-
            Warning code.
        warn_msg:
          type: string
          description: >-
            Human-readable warning message.
    TTSErrorPayload:
      type: object
      properties:
        type:
          type: string
          const: Error
          description: >-
            Message type identifier.
        err_code:
          type: string
          description: >-
            Error code.
        err_msg:
          type: string
          description: >-
            Human-readable error message.