Free
guide
How to caption short videos so they read on mute
Most feed video is watched on mute. A simple model for captions that read well without covering the picture, and how to apply it with the free Captions Overlay Skill.
Nov 3, 2026 · 1 min read
SOFT
Most short video plays on mute, so captions carry the message. Bad captions sit in a thick band, show every "um" and cover the product. Good ones are small, steady and, once in a while, spectacular.
Drop, rail or embed
| Kind | What | How it shows |
|---|---|---|
| Drop | Filler: ums, restarts, self-corrections | Not shown |
| Rail | Ordinary words, verbatim | Small word-by-word line near the bottom, over the picture |
| Embed | The one peak word of a beat | Big, behind the subject, with its own entrance and exit |
Embeds are scarce: at most one per beat, never two on screen together. A 15-second clip usually has one.
Don't reserve a band
The caption line is an overlay. Keep your composition centred on the true middle of the frame and let images run to the bottom edge. Only keep small critical text, such as a URL, out of the strip where the rail sits.
Apply it with Captions Overlay
Captions Overlay is a free, open-source Skill from the HyperFrames project that teaches your agent this model. Ask:
Add captions to this 12-second runner clip: rail captions for the voice-over, and make "soft" the one embed, set behind the runner.For the embed you need the subject on its own layer, such as a product or person cut-out.
Related
For release videos with voice-over and captions built in, see Changelog Video. For a music-led piece with no words to caption, try Motion Reel.
