openapi: 3.0.3
info:
  title: Gladia Control AudioToText File Management API
  description: Gladia AI audio infrastructure API for speech-to-text transcription via REST and WebSocket. Supports asynchronous pre-recorded audio processing and real-time live transcription with speaker diarization, automatic language detection across 100+ languages, and audio intelligence features.
  version: '1.0'
  contact: {}
servers:
- url: https://api.gladia.io/
  description: Gladia API production URL
tags:
- name: File Management
paths:
  /v2/upload:
    post:
      operationId: FileController_upload_v2
      parameters: []
      requestBody:
        required: true
        content:
          multipart/form-data:
            schema:
              type: object
              properties:
                audio:
                  type: string
                  format: binary
                  description: The file to be uploaded. This should be an audio or video file.
          application/json:
            schema:
              type: object
              properties:
                audio_url:
                  type: string
                  description: The URL of the audio or video file to be uploaded.
      responses:
        '200':
          description: ''
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/AudioUploadResponse'
      security:
      - x_gladia_key: []
      summary: Upload an audio file or provide an audio URL for processing
      tags:
      - File Management
components:
  schemas:
    AudioUploadMetadataDTO:
      type: object
      properties:
        id:
          type: string
          description: Uploaded audio file ID
          format: uuid
          example: 6c09400e-23d2-4bd2-be55-96a5ececfa3b
        filename:
          type: string
          description: Uploaded audio filename
          example: short-audio-en-16000.wav
        source:
          type: string
          description: Uploaded audio source
          format: uri
          example: http://files.gladia.io/example/audio-transcription/split_infinity.wav
        extension:
          type: string
          description: Uploaded audio detected extension
          format: uuid
          example: wav
        size:
          type: integer
          description: Uploaded audio size
          example: 365702
        audio_duration:
          type: number
          description: Uploaded audio duration
          example: 4.145782
        number_of_channels:
          type: integer
          description: Uploaded audio channel numbers
          example: 1
      required:
      - id
      - filename
      - extension
      - size
      - audio_duration
      - number_of_channels
    AudioUploadResponse:
      type: object
      properties:
        audio_url:
          type: string
          description: Uploaded audio file Gladia URL
          format: uri
          example: https://api.gladia.io/file/6c09400e-23d2-4bd2-be55-96a5ececfa3b
        audio_metadata:
          description: Uploaded audio file detected metadata
          allOf:
          - $ref: '#/components/schemas/AudioUploadMetadataDTO'
      required:
      - audio_url
      - audio_metadata
  securitySchemes:
    x_gladia_key:
      type: apiKey
      in: header
      name: x-gladia-key
      description: Your personal Gladia API key