OpenAI Whisper

Workflow task type to call OpenAI's Audio Transcription API (speech-to-text model).

Inputs

• file: The audio file to transcribe (required input).
• model: The model to use for transcription (required input).
• language: The language of the audio file. Providing language will improve the accuracy of the transcription. Must be in ISO 639-1 format.
• prompt: Optional prompt to guide the model's style or continue a previous audio segment. Should match the language of the audio file.
• response_format: The format of the output, in one of these options: json, text, srt, verbose_json or vtt.
• temperature: The temperature to use for the model.

Outputs

• text: The transcribed text (default output).

Example Usage

{
  "name": "transcribe_audio",
  "type": "openai.audio.transcription",
  "inputs": {
    "file": "{{inputs.file}}",
    "model": "whisper-1",
    "language": "en",
    "prompt": "The audio is a conversation between two people",
    "response_format": "text",
    "temperature": 0.2
  }
}

This example assumes that the workflow has an input named file that contains an audio file.

Specification

{
  "specification": {
    "type": "openai.audio.transcription",
    "category": "ai",
    "description": "Call OpenAI's Whisper model to transcribe the audio",
    "inputs": [
      {
        "name": "file",
        "type": "file",
        "required": true,
        "description": "The audio file to transcribe",
        "example": "<<file>>"
      },
      {
        "name": "model",
        "type": "text",
        "required": true,
        "default": "whisper-1",
        "description": "The model to use for transcription",
        "example": "whisper-1"
      },
      {
        "name": "language",
        "type": "text",
        "required": false,
        "description": "The language of the audio file. Providing language will improve the accuracy of the transcription",
        "example": "en"
      },
      {
        "name": "prompt",
        "type": "text",
        "required": false,
        "description": "Optional prompt to guide the model's style or continue a previous audio segment. Should match the language of the audio file",
        "example": "The audio is a conversation between two people"
      },
      {
        "name": "response_format",
        "type": "text",
        "required": false,
        "description": "The format of the output, in one of these options: `json`, `text`, `srt`, `verbose_json` or `vtt`",
        "example": "text"
      },
      {
        "name": "temperature",
        "type": "number",
        "required": false,
        "default": 0.0,
        "description": "The temperature to use for the model",
        "example": "0.0"
      }
    ],
    "outputs": [
      {
        "name": "text",
        "type": "text",
        "default": true,
        "description": "The transcribed text",
        "example": "Hello, how are you?"
      }
    ]
  }
}