YouTube's auto-captions miss scripture references, theological terms, and punctuation. Here's the 15-minute workflow to download a YouTube sermon's audio, transcribe it with Whisper accuracy, and end up with a clean, publishable transcript.
Words like 'propitiation', 'eschatology', 'pneumatology', or 'Septuagint' rarely come through correctly. YouTube wasn't trained for theological speech.
YouTube outputs a single wall of caption text with rough punctuation. For a blog post or research notes, you need clean paragraphs, which is what Whisper produces.
You can scrape captions via a transcript panel, but you get raw text with no SRT/VTT/JSON export options. Sermon Transcription gives you all formats with one click.
Total time: 15 minutes for a 45-minute sermon. Your first sermon is free; after that it's covered by the flat $19/month plan.
Copy the full URL from the address bar: youtube.com/watch?v=XXXXX or youtu.be/XXXXX. For your own channel, you can also download direct from YouTube Studio (Content → ... menu → Download).
Option A (recommended): yt-dlp: install with brew install yt-dlp or pip install yt-dlp. Then run: yt-dlp -x --audio-format mp3 --audio-quality 5 URL. Quality 5 = ~96kbps, which keeps even multi-hour recordings under our 400MB cap.
Option B: Browser extension: Free Video Downloader (Firefox) or similar grab the file while you stream. Look for the MP4 option, then convert to MP3 with Audacity.
Option C: Screen record + extract audio: record the video playback with QuickTime (Mac), Game Bar (Windows), or OBS. Export audio with ffmpeg or Audacity.
Check the resulting file size. If it's over 400MB, re-export at lower quality. Quick ffmpeg recipe: ffmpeg -i input.mp3 -b:a 64k -ac 1 output.mp3. A 64kbps mono file is plenty good for transcription accuracy.
Drag and drop the MP3. Whisper auto-detects language. A 45-minute sermon and a two-hour livestream are both included in the flat $19/month plan: useful when you're captioning a whole back catalogue rather than one video.
Within 3-5 minutes you'll have TXT, SRT, VTT, and JSON exports. The TXT version has proper sentence punctuation, paragraph breaks, and accurate scripture references: far cleaner than YouTube's auto-captions.
In YouTube Studio: Subtitles → Add language → English → Upload file → Without timing or With timing (use With timing for the SRT). Publish. Your video now has accurate captions, which improves YouTube ranking and viewer retention.
For one 45-minute YouTube sermon, here's what you'd pay across the popular options.
| Method | Cost / sermon | Turnaround | Output quality |
|---|---|---|---|
| YouTube auto-captions | Free | Minutes after upload | ~75-85% accuracy, no formatting |
| Sermon Transcription (this site) | $19/mo flat, unlimited (first sermon free) | 5 minutes | 99% accuracy, paragraph-clean |
| Otter.ai (Business) | $20-30/mo subscription | Real-time | ~90% accuracy, monthly minute cap |
| Rev.com (AI) | $11.25 (at $0.25/min) | Same day | ~95% accuracy |
| Rev.com (Human) | $67.50 (at $1.50/min) | 12-24 hours | 99%+ accuracy |
For sermon use cases (where 95-98% machine accuracy + 10 min of editing is functionally identical to human work), AI transcription is the obvious winner.
Three steps, 15 minutes, covered by the flat plan.
Otter has Chrome extensions for YouTube. Here's how it compares head-to-head.
TurboScribe is another Whisper wrapper. Pricing and feature differences.
Detailed walkthrough on uploading SRTs to YouTube for accurate captions.
The full updated guide covering all sources, including YouTube.
Download, upload, get a transcript in 15 minutes. First sermon free.
Start free