Captions
How do I add captions to a video automatically?
Short answer
Run the video through speech-to-text to get word-level timestamps, then render those words onto the video as burned-in captions. Platform auto-captions exist but are styled by the platform and can be turned off by viewers; burned-in captions always show and are what short-form creators use.
Step by step
- 1
Transcribe with word-level timing
Sentence-level timing is enough for a subtitle file but not for the word-by-word style short-form uses. Make sure the tool returns per-word timestamps.
- 2
Fix the names
Speech-to-text reliably mangles proper nouns, product names, and jargon. This is the one manual pass worth doing — a misspelled name in giant text is what people screenshot.
- 3
Choose a readable style
Heavy sans-serif, large, high contrast, with an outline or shadow so it survives a bright background. Style is branding; legibility is reach.
- 4
Position inside the safe area
Centre-ish vertically, clear of the bottom third where platform UI sits. Captions hidden behind a username are captions nobody read.
- 5
Burn them in on export
Rendered into the pixels, not attached as a sidecar file. Uploaded .srt files are viewer-toggleable and default off in most feeds.
The detail
The reason this matters more than it sounds: a large share of short-form watch time happens with sound off, in public, at speed. Captions aren't an accessibility afterthought in this format — they are how most of the audience receives the content at all.
Word-by-word or phrase-by-phrase highlighting outperforms static blocks of text because the motion itself holds attention. That's also why it's overused: if every clip on the feed has bouncing neon words, style stops differentiating you and only legibility matters.
Keep a separate .srt if you also publish long-form — YouTube uses uploaded subtitle files for search indexing, and that's real discovery value that burned-in captions don't provide.
Follow-up questions
Are platform auto-captions good enough?
For accuracy, often yes. For performance, no — they're styled by the platform, positioned unpredictably, and viewers can switch them off.
How accurate is automatic transcription?
Very high for clear speech in a well-recorded room, and noticeably worse with accents, crosstalk, background noise, or specialist vocabulary. Always read it back before exporting.
Do captions really improve watch time?
Consistently, in this format. It's the single change with the most reliable effect on retention for talking-head and narration content.
Also asked as: auto caption a video · add subtitles to video automatically free · burn captions into a video