Inworld TTS API

openapi: 3.1.0
info:
  title: Inworld TTS API
  description: >
    Inworld Text-to-Speech API. Synchronous synthesis (returns base64 audio in one
    response), server-streamed synthesis (chunked audio for low-latency playback),
    and WebSocket-based bidirectional streaming synthesis. Backed by the Realtime
    TTS-2, Realtime TTS 1.5 Max, and Realtime TTS 1.5 Mini models with word- and
    character-level alignment, phoneme/viseme metadata for lipsync, custom
    pronunciation, pause controls, voice tags, and long-text input.
  version: v1
  contact:
    name: Inworld Support
    url: https://docs.inworld.ai/tts/resources/support
  license:
    name: Inworld Terms of Service
    url: https://inworld.ai/legal/terms-of-service
servers:
  - url: https://api.inworld.ai
    description: Inworld Production API
security:
  - BasicAuth: []
tags:
  - name: Text To Speech
    description: Synthesize speech from text using Inworld voice models.
paths:
  /tts/v1/voice:
    post:
      summary: Synthesize Speech
      description: >
        Synthesize a single audio response from text. Returns base64-encoded audio
        plus optional word/character/phoneme alignment metadata. Maximum 2,000 input
        characters per request and 16 MB output audio.
      operationId: synthesizeSpeech
      tags:
        - Text To Speech
      requestBody:
        required: true
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/SynthesizeRequest'
            examples:
              Default:
                $ref: '#/components/examples/SynthesizeExample'
      responses:
        '200':
          description: Synthesized audio returned successfully.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/SynthesizeResponse'
        '4XX':
          $ref: '#/components/responses/ErrorResponse'
  /tts/v1/voice:stream:
    post:
      summary: Stream Synthesize Speech
      description: >
        Stream synthesis. The response is a chunked sequence of JSON envelopes,
        each containing a base64 audio fragment. Use for low-latency playback of
        long-form text.
      operationId: streamSynthesizeSpeech
      tags:
        - Text To Speech
      requestBody:
        required: true
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/SynthesizeRequest'
      responses:
        '200':
          description: Streamed audio chunks.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/SynthesizeStreamChunk'
        '4XX':
          $ref: '#/components/responses/ErrorResponse'
  /tts/v1/voice:websocket:
    get:
      summary: WebSocket Synthesize Speech
      description: >
        Upgrade to a WebSocket connection for bidirectional streaming synthesis.
        Clients send `SynthesizeRequest` frames and receive chunked audio frames
        with timestamp metadata. Auth via `Authorization: Basic` header on the
        upgrade request.
      operationId: websocketSynthesizeSpeech
      tags:
        - Text To Speech
      responses:
        '101':
          description: Switching protocols to WebSocket.
  /tts/v1/voices:
    get:
      summary: List TTS Voices
      description: >
        List voices available for synthesis. Includes Inworld-shipped voices and
        any custom-cloned or designed voices in the caller's workspace.
      operationId: listTtsVoices
      tags:
        - Text To Speech
      parameters:
        - name: filter
          in: query
          required: false
          description: Optional filter expression (e.g. by tag, language, gender).
          schema:
            type: string
      responses:
        '200':
          description: Voices returned successfully.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/ListVoicesResponse'
        '4XX':
          $ref: '#/components/responses/ErrorResponse'
components:
  securitySchemes:
    BasicAuth:
      type: http
      scheme: basic
      description: >
        Inworld API key. The Inworld portal issues a value already pre-encoded as
        `base64(apiKey:)` — send it as `Authorization: Basic <api-key>`.
  schemas:
    SynthesizeRequest:
      type: object
      required:
        - text
        - voiceId
        - modelId
        - audioConfig
      properties:
        text:
          type: string
          maxLength: 2000
          description: Text to synthesize. Maximum 2,000 characters per request.
        voiceId:
          type: string
          description: Voice identifier (shipped voice name or custom voice id).
        modelId:
          type: string
          description: Inworld TTS model id.
          enum:
            - inworld-tts-2
            - inworld-tts-1.5-max
            - inworld-tts-1.5-mini
        audioConfig:
          $ref: '#/components/schemas/AudioConfig'
        language:
          type: string
          description: BCP-47 language tag (e.g. `en-US`).
        deliveryMode:
          type: string
          description: TTS-2 only. Controls stability vs. expressiveness.
          enum: [STABLE, BALANCED, CREATIVE]
          default: BALANCED
        temperature:
          type: number
          minimum: 0
          maximum: 2
          default: 1.0
          description: Randomness. Ignored on TTS-2.
        timestampType:
          type: string
          enum: [WORD, CHARACTER]
          description: If set, return alignment metadata of the requested granularity.
        applyTextNormalization:
          type: string
          enum: [ON, OFF]
          description: Normalize numbers, dates, abbreviations, etc. before synthesis.
    AudioConfig:
      type: object
      required:
        - audioEncoding
      properties:
        audioEncoding:
          type: string
          enum: [LINEAR16, MP3, OGG_OPUS, ALAW, MULAW, FLAC, PCM, WAV]
        sampleRateHertz:
          type: integer
          minimum: 8000
          maximum: 48000
          default: 48000
        bitRate:
          type: integer
          default: 128000
          description: Bits per second. Compressed formats only.
        speakingRate:
          type: number
          minimum: 0.5
          maximum: 1.5
          default: 1.0
    SynthesizeResponse:
      type: object
      properties:
        audioContent:
          type: string
          format: byte
          description: Base64-encoded audio bytes.
        usage:
          $ref: '#/components/schemas/SynthesisUsage'
        timestampInfo:
          $ref: '#/components/schemas/TimestampInfo'
    SynthesizeStreamChunk:
      type: object
      properties:
        audioContent:
          type: string
          format: byte
        timestampInfo:
          $ref: '#/components/schemas/TimestampInfo'
        usage:
          $ref: '#/components/schemas/SynthesisUsage'
    SynthesisUsage:
      type: object
      properties:
        processedCharactersCount:
          type: integer
        modelId:
          type: string
    TimestampInfo:
      type: object
      properties:
        wordAlignment:
          $ref: '#/components/schemas/WordAlignment'
        characterAlignment:
          $ref: '#/components/schemas/CharacterAlignment'
    WordAlignment:
      type: object
      properties:
        words:
          type: array
          items:
            type: string
        wordStartTimeSeconds:
          type: array
          items:
            type: number
        wordEndTimeSeconds:
          type: array
          items:
            type: number
        phoneticDetails:
          type: array
          items:
            $ref: '#/components/schemas/PhoneticDetail'
    PhoneticDetail:
      type: object
      properties:
        wordIndex:
          type: integer
        phones:
          type: array
          items:
            type: object
            properties:
              phoneSymbol:
                type: string
                description: IPA phone symbol.
              startTimeSeconds:
                type: number
              durationSeconds:
                type: number
              visemeSymbol:
                type: string
                description: Viseme label for lipsync.
        isPartial:
          type: boolean
    CharacterAlignment:
      type: object
      properties:
        characters:
          type: array
          items:
            type: string
        characterStartTimeSeconds:
          type: array
          items:
            type: number
        characterEndTimeSeconds:
          type: array
          items:
            type: number
    ListVoicesResponse:
      type: object
      properties:
        voices:
          type: array
          items:
            $ref: '#/components/schemas/Voice'
    Voice:
      type: object
      properties:
        voiceId:
          type: string
        displayName:
          type: string
        languages:
          type: array
          items:
            type: string
        gender:
          type: string
        tags:
          type: array
          items:
            type: string
        source:
          type: string
          enum: [INWORLD, IVC, DESIGN]
    Error:
      type: object
      properties:
        code:
          type: integer
        message:
          type: string
        details:
          type: array
          items:
            type: object
  responses:
    ErrorResponse:
      description: Error response.
      content:
        application/json:
          schema:
            $ref: '#/components/schemas/Error'
  examples:
    SynthesizeExample:
      summary: Basic synthesis with LINEAR16
      value:
        text: "Hello, world! What a wonderful day to be a text-to-speech model."
        voiceId: Dennis
        modelId: inworld-tts-2
        audioConfig:
          audioEncoding: LINEAR16
          sampleRateHertz: 22050
        deliveryMode: BALANCED
        applyTextNormalization: ON