Guide
How to Add Captions to Video Ads (Styles, SRT, Safe Zones)
Why captions matter for video ads, how to pick a style, where to place them on TikTok, Reels and Shorts, and how to burn them in or export an SRT.
· 5 min read
A large share of social video is watched with the sound off, and even viewers with sound on read faster than they listen. Captions make the message land either way, keep attention on the screen, and make your ads usable for viewers who are deaf or hard of hearing. For ads, they are not optional.
This guide covers the two ways to caption, how to choose between the five styles in AI Captions, and where captions should sit so platform buttons do not cover them.
Burned-in captions vs an SRT file
| Burned-in (open) captions | SRT (closed) captions | |
|---|---|---|
| What it is | Text drawn into the video frames | A separate text file with timestamps |
| Always visible | Yes | Only if the player shows it and the viewer turns it on |
| Styling | Full control: font, color, highlight | Decided by the player |
| Edit later | Re-export the video | Edit the text file |
| Best for | TikTok, Reels, Shorts, feed ads | YouTube, websites, accessibility uploads |
AI Captions gives you both from one upload: a new MP4 with the captions burned in, and an SRT file of the same transcript. One run costs 25 credits.
How it works
- Upload an MP4. The audio is transcribed with Whisper (up to 3 minutes of audio per video, which covers almost every ad).
- Pick a style and whether you want the text in uppercase.
- The captions are timed to the speech and burned into a new MP4. You download the video and the SRT.
Proofread the result. Speech recognition is good but not perfect, especially with brand names, product codes and strong accents. Fix the SRT text if needed; for the burned-in version, a clear voice track and a pronounceable brand name help most.
The five styles
| Style | Look | Use it for |
|---|---|---|
bold-pop | Big heavy type with a thick outline, short word groups | UGC and talking-head ads on TikTok and Reels |
clean | Simple sans-serif with a soft shadow | Premium brands, explainers, B2B and LinkedIn |
boxed | Text on a solid rounded box | Busy or bright footage where outlines are not enough |
neon | Glowing colored text | Night scenes, gaming, music, nightlife |
karaoke | Each word highlights as it is spoken | Fast talkers, tutorials, keeping eyes moving |
If you are unsure, start with bold-pop for creator content and clean for polished brand video. Use boxed whenever you can barely read a test frame.
Placement and safe zones
On vertical platforms, the app's own interface covers parts of the frame: the caption and username at the bottom, the action buttons on the right, and sometimes a top bar. Exact areas vary by app and change over time, so check each platform's current guidance, but these rules hold up:
- Keep captions out of roughly the bottom fifth of a 9:16 frame, where the post caption and buttons sit.
- Keep text away from the right edge, where like, comment and share buttons stack.
- Avoid the very top strip, which can be covered by the app header or ad label.
- The safest band is the middle of the frame, slightly below center.
- For 4:5 feed video, the interface overlaps less, but leave a margin at the bottom for the CTA button.
Plan this during the shoot: leave space in the composition where captions will go, and keep the creator's face out of that band. The ad specs by platform guide has the sizes to export.
Caption writing tips
- Short lines: two to five words on screen at a time reads faster than full sentences.
- Caption the hook even if nothing else: the first line has to work silently.
- Uppercase is louder and suits energetic ads; sentence case is calmer and easier to read in long passages.
- Do not stack captions on top of other on-screen text. If you add a headline, move one of them.
- Keep brand and product names spelled consistently; fix them in the transcript before export if needed.
Captions in a full workflow
If your ad uses an AI voice instead of a recorded one, generate the voiceover first, lay it under the edit, export, then caption the final cut so the timing matches what viewers hear. For many ads at once, the captions endpoint in the API takes an MP4 and a style and returns the captioned video and SRT, which is easy to wire into n8n or Make.
Finally, run the captioned version through the Ad Analyzer. It reads the key frames and transcript and will flag a hook that does not come across without sound.