Video Transcriber Actor 🎤🎬

This Apify Actor automates the process of downloading videos from public URLs, extracting their audio content, and then transcribing the audio into text using OpenAI's powerful speech-to-text models (GPT-4o Mini Transcribe or GPT-4o Transcribe).

Features

• Batch Video Processing: Provide a list of video URLs and get transcriptions for all of them.
• Powered by OpenAI: Utilizes state-of-the-art AI models (gpt-4o-mini-transcribe or gpt-4o-transcribe) for accurate transcriptions.
• Configurable Transcription: Adjust settings like language, prompt, and temperature to fine-tune transcription results.
• Robust Error Handling: Implements retries for network issues or temporary API failures.
• Parallel Processing: Downloads and transcribes multiple videos concurrently for faster results.
• Secure API Key Handling: Your OpenAI API key is treated as a secret input.

Use Cases

• Transcribing lectures, talks, or presentations.
• Generating subtitles or text content from video podcasts.
• Making video content searchable by transcribing its audio.
• Analyzing spoken content in a collection of videos.

Input Configuration

The actor requires the following input fields. Your OpenAI API key is essential for the transcription service to work.

Field	Type	Description	Default Value
video_urls	Array	Required. A list of public direct URLs to video files (e.g., MP4, MOV, AVI). Each URL will be processed.	[] (Example prefilled)
openai_api_key	String	Required. Your OpenAI API key. This is treated as a secret and stored securely.	N/A
openai_model	String	The OpenAI model for transcription. gpt-4o-mini-transcribe is fast & cost-effective; gpt-4o-transcribe may offer higher accuracy.	gpt-4o-mini-transcribe
openai_transcription_language	String	Optional. Language of the audio in ISO-639-1 format (e.g., en for English). If omitted, OpenAI attempts auto-detection.	"" (Empty String)
openai_transcription_prompt	String	Optional. Text prompt to guide the model's style or vocabulary (e.g., for specific jargon or names).	N/A
openai_transcription_temperature	String	Sampling temperature (0.0-1.0, provided as a string e.g., "0.2"). Lower values are more deterministic.	"0.0"
max_concurrent_tasks	Integer	Maximum number of videos to process in parallel.	5
max_retries	Integer	Number of times to retry processing a video if an error occurs.	3

Example Input JSON:

{
    "video_urls": [
        "https://www.ffmpeg.org/example-assets/Counting_Atoms_preview.mp4",
        "https://another-public-domain.com/another-video.mp4"
    ],
    "openai_api_key": "sk-yourSecretOpenAiApiKeyGoesHere",
    "openai_model": "gpt-4o-mini-transcribe",
    "openai_transcription_language": "en",
    "openai_transcription_prompt": "Focus on scientific terminology.",
    "openai_transcription_temperature": "0.2",
    "max_concurrent_tasks": 5,
    "max_retries": 3
}

Output

The actor saves each transcription result as a separate item in the Apify Dataset. Each item will have the following structure:

{
  "download_url": "https://www.example.com/video.mp4",
  "transcription": "This is the transcribed text from the video...",
  "status": "succeeded" // or "failed"
}

If a video fails to process after all retries, the transcription will be null, status will be failed, and an error field will contain the error message.

How to Use

1. Go to the Actor page on the Apify Store.
2. Click on "Try actor".
3. Fill in the input configuration fields, especially video_urls and your openai_api_key.
4. Click "Start" to run the actor.
5. When the run finishes, you can find the results in the "Dataset" tab of the run console.

Technical Details

• The actor uses ffmpeg to extract audio from video files. Ensure the video formats are compatible with common ffmpeg builds.
• Video downloads are performed asynchronously.
• Transcription tasks are processed in parallel using Python's multiprocessing.

Limitations

• URL Accessibility: Video URLs must be publicly accessible and direct links to video files. Redirects are followed, but complex authentication or sites requiring browser interaction are not supported.
• OpenAI API Limits: Your OpenAI API usage is subject to your OpenAI account's rate limits and quotas. Long videos or large batches might take time or hit these limits.
• Video Size/Length: Extremely large video files might lead to increased processing time or memory usage. The actor downloads the entire video into memory before audio extraction.
• CDN Link Stability: If using temporary CDN links (e.g., from some social media platforms), they may expire. Prefer stable, direct URLs.

Support & Issues

If you encounter any issues or have suggestions for improvement, please open an issue on the GitHub repository for this actor (if applicable, or provide another contact method).

---

Happy Transcribing!