YouTube Transcriber

One-command transcription for any YouTube video — even when subtitles are disabled.

How it works:

1. Tries to fetch existing YouTube subtitles first (free, instant)
2. If subtitles are disabled/unavailable → downloads audio → transcribes via OpenAI Whisper API

Prerequisites

• yt-dlp — brew install yt-dlp or pip install yt-dlp
• ffmpeg — brew install ffmpeg
• OPENAI_API_KEY environment variable set

Usage

# Basic — auto-detect best method
{baseDir}/scripts/transcribe.sh "https://www.youtube.com/watch?v=VIDEO_ID"

# Specify language (improves accuracy for non-English)
{baseDir}/scripts/transcribe.sh "https://www.youtube.com/watch?v=VIDEO_ID" --lang zh

# Custom output path
{baseDir}/scripts/transcribe.sh "https://www.youtube.com/watch?v=VIDEO_ID" --out /tmp/transcript.txt

# Force Whisper (skip subtitle check)
{baseDir}/scripts/transcribe.sh "https://www.youtube.com/watch?v=VIDEO_ID" --force-whisper

# Keep downloaded audio file
{baseDir}/scripts/transcribe.sh "https://www.youtube.com/watch?v=VIDEO_ID" --keep-audio

Options

How It Works

YouTube URL
    │
    ├─► Try yt-dlp subtitles (free, instant)
    │       │
    │       ├─ Found → Clean & output transcript ✓
    │       │
    │       └─ Not found / disabled
    │               │
    │               ▼
    └─► Download audio (yt-dlp + ffmpeg)
            │
            ├─ Compress to mono, low bitrate (fit 25MB API limit)
            │
            ▼
        OpenAI Whisper API → transcript ✓

Cost

• With subtitles available: Free (uses existing captions)
• Whisper API: ~$0.006/minute of audio → 23-min video ≈ $0.14

Supported

• Any YouTube video URL (full URL, youtu.be short link, or video ID)
• All languages supported by Whisper (99+ languages)
• Videos up to ~4 hours (audio compressed to fit 25MB API limit)

Troubleshooting

• 403 from YouTube: Update yt-dlp (pip install -U yt-dlp)
• File too large for Whisper: Lower bitrate with --audio-bitrate 32
• Missing API key: Set OPENAI_API_KEY environment variable