openapi: 3.1.0
info:
  title: Fish Audio Asr Tts API
  description: Fish Audio is an AI voice platform offering text-to-speech, voice cloning, speech-to-text, and voice model management. Authentication is by API key issued at https://fish.audio/app/api-keys and presented as a Bearer token.
  version: '1.0'
  contact:
    name: Fish Audio
    url: https://docs.fish.audio
servers:
- url: https://api.fish.audio
  description: Production
security:
- bearerAuth: []
tags:
- name: Tts
paths:
  /v1/tts:
    post:
      summary: Generate speech from text
      operationId: textToSpeech
      parameters:
      - name: model
        in: header
        required: false
        description: TTS model variant (s1 or s2-pro)
        schema:
          type: string
          enum:
          - s1
          - s2-pro
      requestBody:
        required: true
        content:
          application/json:
            schema:
              type: object
              required:
              - text
              properties:
                text:
                  type: string
                reference_id:
                  type: string
                  description: ID of a voice model to use
                format:
                  type: string
                  enum:
                  - mp3
                  - wav
                  - pcm
                  - opus
                  default: mp3
                latency:
                  type: string
                  enum:
                  - normal
                  - balanced
                prosody:
                  type: object
                  properties:
                    speed:
                      type: number
                    volume:
                      type: number
                chunk_length:
                  type: integer
      responses:
        '200':
          description: Audio stream
          content:
            audio/mpeg:
              schema:
                type: string
                format: binary
      tags:
      - Tts
  /v1/tts/stream/with-timestamp:
    post:
      summary: Stream TTS with per-chunk timestamps (Server-Sent Events)
      operationId: textToSpeechStreamWithTimestamp
      parameters:
      - name: model
        in: header
        required: false
        schema:
          type: string
          enum:
          - s1
          - s2-pro
      requestBody:
        required: true
        content:
          application/json:
            schema:
              type: object
              required:
              - text
              properties:
                text:
                  type: string
                reference_id:
                  type: string
                chunk_length:
                  type: integer
      responses:
        '200':
          description: SSE stream of audio chunks plus timing metadata
          content:
            text/event-stream:
              schema:
                type: string
      tags:
      - Tts
components:
  securitySchemes:
    bearerAuth:
      type: http
      scheme: bearer