Inworld STT API

openapi: 3.1.0
info:
  title: Inworld STT API
  description: >
    Inworld Speech-to-Text. Synchronous transcription and a streaming WebSocket
    endpoint. Multi-provider routing (provider/model id, e.g. `groq/whisper-large-v3`),
    99+ languages, word timestamps, contextual prompts, end-of-turn detection, and
    voice-profile-aware transcription for conversational agents.
  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: Speech To Text
    description: Transcribe audio to text.
paths:
  /stt/v1/transcribe:
    post:
      summary: Transcribe Audio
      description: >
        Transcribe a base64-encoded audio payload to text. Optionally returns word
        timestamps and routes through the requested provider/model id.
      operationId: transcribeAudio
      tags: [Speech To Text]
      requestBody:
        required: true
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/TranscribeRequest'
            examples:
              Default:
                $ref: '#/components/examples/TranscribeExample'
      responses:
        '200':
          description: Transcription returned.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/TranscribeResponse'
        '4XX':
          $ref: '#/components/responses/ErrorResponse'
  /stt/v1/transcribe:stream:
    get:
      summary: Streaming Transcribe (WebSocket)
      description: >
        Upgrade to a WebSocket. Send PCM/Opus audio chunks and receive interim and
        final transcripts with word timestamps. Auth via `Authorization: Basic` on
        the upgrade.
      operationId: streamTranscribe
      tags: [Speech To Text]
      responses:
        '101':
          description: Switching protocols to WebSocket.
components:
  securitySchemes:
    BasicAuth:
      type: http
      scheme: basic
  schemas:
    TranscribeRequest:
      type: object
      required: [transcribeConfig, audioData]
      properties:
        transcribeConfig:
          $ref: '#/components/schemas/TranscribeConfig'
        audioData:
          type: object
          required: [content]
          properties:
            content:
              type: string
              format: byte
              description: Base64-encoded audio in the configured encoding.
    TranscribeConfig:
      type: object
      required: [modelId, audioEncoding]
      properties:
        modelId:
          type: string
          description: provider/model identifier, e.g. `groq/whisper-large-v3`.
        audioEncoding:
          type: string
          enum: [LINEAR16, MP3, OGG_OPUS, FLAC, AUTO_DETECT]
        language:
          type: string
          description: BCP-47 language tag.
        sampleRateHertz:
          type: integer
          default: 16000
        numberOfChannels:
          type: integer
          default: 1
        inactivityTimeoutSeconds:
          type: integer
        endOfTurnConfidenceThreshold:
          type: number
          minimum: 0
          maximum: 1
          default: 0.5
        prompts:
          type: array
          items:
            type: string
          description: Contextual hint strings to bias recognition.
        includeWordTimestamps:
          type: boolean
        voiceProfileConfig:
          type: object
          properties:
            enabled:
              type: boolean
            profileId:
              type: string
    TranscribeResponse:
      type: object
      properties:
        transcription:
          type: object
          properties:
            transcript:
              type: string
            isFinal:
              type: boolean
            wordTimestamps:
              type: array
              items:
                type: object
                properties:
                  word:
                    type: string
                  startTimeSeconds:
                    type: number
                  endTimeSeconds:
                    type: number
                  confidence:
                    type: number
        usage:
          type: object
          properties:
            transcribedAudioMs:
              type: integer
            modelId:
              type: string
    Error:
      type: object
      properties:
        code:
          type: integer
        message:
          type: string
  responses:
    ErrorResponse:
      description: Error response.
      content:
        application/json:
          schema:
            $ref: '#/components/schemas/Error'
  examples:
    TranscribeExample:
      summary: Transcribe English audio via Whisper Large v3 on Groq.
      value:
        transcribeConfig:
          modelId: groq/whisper-large-v3
          audioEncoding: LINEAR16
          language: en-US
          sampleRateHertz: 16000
          numberOfChannels: 1
        audioData:
          content: "<base64-encoded-audio>"