openapi: 3.2.0
info:
  title: Sakura Internet Audio API
  version: 1.0.1
  description: 'Operations tagged Audio across 2 of this provider''s published API definitions: ai-engine-inference-api.yaml, sakura-internet-ai-engine-inference-openapi.yml. Each path carries the servers of the definition it was published in.'
servers:
- url: https://api.ai.sakura.ad.jp
security:
- BearerAuth: []
tags:
- name: Audio
paths:
  /v1/audio/transcriptions:
    post:
      summary: Create a transcription
      operationId: createTranscription
      requestBody:
        required: true
        content:
          multipart/form-data:
            schema:
              type: object
              required:
              - file
              properties:
                file:
                  type: string
                  format: binary
                  description: 'Audio file to transcribe. Common formats: aac, m4a, mp3, mp4, ogg, wav etc.

                    '
                model:
                  type: string
                  description: Transcription model identifier served by vLLM.
                  enum:
                  - whisper-large-v3-turbo
                language:
                  type: string
                  default: ja
                  description: 'Source language hint (BCP-47, e.g. "ja", "en-US").

                    '
                prompt:
                  type: string
                  description: Optional decoding/prompt bias (proper nouns, style hints).
                temperature:
                  type: number
                  minimum: 0
                  maximum: 1
                  default: 0
                  description: Decoding temperature.
                stream:
                  type: boolean
                  default: false
      responses:
        '200':
          description: Transcription created
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/TranscriptionResponse'
              examples:
                exampleOk:
                  summary: Successful transcription
                  value:
                    model: whisper-large-v3-turbo
                    text: 本日はご利用いただきありがとうございます。
        '400':
          description: Bad request
        '401':
          description: Unauthorized
        '429':
          description: Rate limited
        '500':
          description: Server error
        '504':
          description: Server error
      tags:
      - Audio
    servers:
    - url: https://api.ai.sakura.ad.jp
  /v1/audio/speech:
    post:
      summary: Create speech (text-to-speech)
      operationId: createSpeech
      description: 'テキストから音声を生成します（TTS）。

        - 必須: input, model

        - instructionsは指定できますが現在は無視されます

        - response_formatは指定できますが現在は常にwavを返します

        - streamは非対応です（stream_formatを指定してもストリーミングにはなりません）'
      requestBody:
        required: true
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/SpeechRequest'
            examples:
              exampleWav:
                summary: Generate wav speech (response_format is ignored)
                value:
                  model: zundamon
                  voice: normal
                  input: こんにちは。
                  instructions: 落ち着いたトーンで話して
                  response_format: wav
                  stream_format: sse
      responses:
        '200':
          description: Speech audio generated (always wav)
          headers:
            Content-Type:
              description: Always `audio/wav`.
              schema:
                type: string
                example: audio/wav
          content:
            audio/wav:
              schema:
                type: string
                format: binary
        '400':
          description: Bad request
        '401':
          description: Unauthorized
        '429':
          description: Rate limited
        '500':
          description: Server error
        '504':
          description: Server error
      tags:
      - Audio
    servers:
    - url: https://api.ai.sakura.ad.jp
components:
  schemas:
    TranscriptionResponse:
      type: object
      required:
      - model
      - text
      properties:
        model:
          type: string
          description: Model used.
        text:
          type: string
          description: Full transcription (plain text).
    SpeechRequest:
      type: object
      required:
      - model
      - input
      properties:
        model:
          type: string
          description: '音声合成モデル識別子(例：zundamon)

            利用可能なmodelはコントロールパネル等をご確認ください。

            '
        input:
          type: string
          description: 音声合成するテキスト（最大1000文字程度）
          minLength: 1
          maxLength: 1000
        voice:
          type: string
          description: '話者/スタイル(例：normal)

            利用可能なvoiceはコントロールパネル等をご確認ください。

            '
        instructions:
          type: string
          description: '追加指示(例： 話し方のトーンなど)。

            ※現在は指定できますが無視されます。

            '
        response_format:
          type: string
          description: '出力フォーマット。

            ※現在は指定できますが常にwavを返します。

            '
          default: wav
          enum:
          - wav
          - mp3
          - ogg
          - aac
          - flac
        stream_format:
          type: string
          description: 'ストリーム形式。

            ※現在は指定できますが無視されます。

            '
          enum:
          - sse
          - jsonl
  securitySchemes:
    BearerAuth:
      type: http
      scheme: bearer
x-refined-from:
- ai-engine-inference-api.yaml
- sakura-internet-ai-engine-inference-openapi.yml