# How do I add captions to a video automatically?

**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.

*Captions*

*Also asked as: auto caption a video; add subtitles to video automatically free; burn captions into a video*

Canonical: https://novaclipper.com/answers/how-do-i-add-captions-to-a-video-automatically

## 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.

## 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.

## FAQ

### 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.

## Related

- [Captions, defined](https://novaclipper.com/glossary/captions)
- [Auto-captioning guide](https://novaclipper.com/blog/how-to-add-captions-to-videos-automatically)
- [Free: download subtitles as SRT/VTT](https://novaclipper.com/tools/subtitle-downloader)

---

Nova turns long-form video into vertical short-form clips with AI moment detection, face-aware 9:16 reframing, and word-by-word captions — and makes faceless videos from a script or a Reddit thread. Free tier: 60 minutes/month, no card. [Start free](https://novaclipper.com/signup) · [All use cases](https://novaclipper.com/for) · [Free tools](https://novaclipper.com/tools)