Audio & Video to Text Transcription

Overview

An Apify actor that transcribes audio and video files into text and subtitle formats using OpenAI's Whisper model. It processes direct URLs to media files (like MP4s) and outputs multiple standard formats.

Key Features

• OpenAI Whisper Integration: Leverages Whisper for accurate speech-to-text transcription.
• Multiple Output Formats: Returns transcriptions in JSON, plain text (TXT), SRT, VTT, and TSV formats in a single run.
• Pre-installed Models: The tiny, base, and small Whisper models are included in the Docker image for faster, offline processing.
• Direct URL Processing: Accepts a publicly accessible URL to an audio or video file.

Input

Configure the actor using a JSON input object with these parameters:

• model (string): The Whisper model to use.
  • Options: tiny, base, small, medium, large, turbo.
  • The tiny, base, and small models are pre-installed for immediate use. Larger models (medium, large, turbo) will be downloaded on first run.
• source_url (string): The direct URL to the video or audio file (e.g., https://example.com/video.mp4).
  • Note: YouTube links are not supported directly. You must provide a URL to a downloadable media file.

Example Input

{
  "model": "small",
  "source_url": "https://example.com/path/to/your/audio.mp3"
}

Output

The actor returns a JSON array containing a single object. This object holds the transcription in five different formats under the following keys:

• json: The full, structured Whisper output including segments, tokens, and metadata.
• txt: A plain text version of the transcribed speech.
• srt: SubRip subtitle format.
• vtt: WebVTT subtitle format.
• tsv: Tab-separated values with columns for start time, end time, and text.

Example Output (Excerpt)

[
  {
    "json": "{ \"text\": \"What's your favorite drink?\", \"segments\": [...] }",
    "txt": "What's your favorite drink?\nMy favorite drink is apple juice.\n",
    "srt": "1\n00:00:00,000 --> 00:00:01,120\nWhat's your favorite drink?\n",
    "vtt": "WEBVTT\n\n00:00.000 --> 00:01.120\nWhat's your favorite drink?",
    "tsv": "start\tend\ttext\n0\t1120\tWhat's your favorite drink?"
  }
]

How to Use

1. In your Apify dashboard, create a new actor or task.
2. Copy and paste the provided actor source code into the editor.
3. Set the input configuration using the JSON structure shown above.
4. Run the actor. It will fetch the media from the provided URL, transcribe it using the specified Whisper model, and return the results.

Important Notes

• This script is provided "as is" without warranty. Use it at your own risk.
• You are responsible for ensuring compliance with relevant terms of service (e.g., YouTube's ToS if you transcribe downloaded content) and copyright laws.
• For transcribing YouTube videos, you must first download the video as an MP4 (or similar) file and host it at a publicly accessible URL.