How to Add Burned-In Captions That Don't Cover the Hands

Anyone cutting a product demo can add burned-in captions in 10 minutes that do not cover the hands: lock a safe band, keep the click readable, burn once. Do it in CapCut, Descript, or Runway before a "make the captions pop" sticker sits on the finger.

Burned-in captions are speech on the picture, not a second hero. If the viewer cannot see which button they pressed, you never captioned. You decorated.

This is not a name lower-third that does not cover the face. A lower-third is a name. This skill is the words they said. It is not a talking-head into a 15-second hook. A hook is a cut. Captions live on the take. It is not soft b-roll under a screen-recording narration. B-roll is a bed. This is type on the A-roll. Save the band lock as a skill prompt you can rerun.

Add burned-in captions that miss the hands

1. Mark the hands. Name the caption band

Play the take once with a pad and a pen. Mark every beat a hand, cursor, or tool enters the lower third: first tap, first fold, first pour, first stitch.

Write three facts on paper: where the hands live (lower third, left, or right), the Y band you will allow for type (top 20 percent, or a side column), and a max of two lines per cue.

If you cannot name those three, watch again. A default caption preset is not a plan.

Banned: auto-captions as a personality, bottom-center forever, karaoke bounce, cover the demo, huge yellow box, make them pop, outline so thick it is a brick, emoji on every line, burn over the hands, match TikTok even when the hands die.

2. Write short captions. Do not write a transcript wall

Instruct ChatGPT, Claude, or Gemini if you need a transcript first: "Two lines max per cue. Cut um and so. Keep their nouns. Do not write a paragraph. If a beat has no speech, write no cue."

Then place in CapCut, Descript, or Runway. Do not paste the whole transcript as one card.

Good cue: "Tap Export, then March." Bad: "So then what you want to do here is go ahead and tap export for the month." That is a novel on a finger.

If a line is long because the speaker rammed two thoughts, split the cue. Two thoughts on one burned-in captions card is a paragraph.

3. Place the band away from the demo. Burn once

Lock a Y position. Top band for a table demo. Side column for a vertical phone. Never the default bottom-center if the hands live there.

Instruct yourself: "Type supports the VO. Hands carry the proof. If I mute the VO and the caption still covers the tap, the band is wrong."

Good band: two lines, top 18 percent, 8 percent margin from the frame edge, contrast that reads on a phone. Bad band: a yellow box sitting on the needle.

If the punchiest word lands on a tap, move the line up, split it, or hold the previous line for a beat. Do not cover the tap.

Hard cap: one style. One size. One color. No karaoke bounce. No second font for "emphasis."

4. Force a mute check on the hands

Export a pass. Mute the VO. Watch only the hands and the type.

Tell yourself three checks:

• Every tap, fold, or pour is still visible.
• No cue sits on a tool for more than a frame of contact.
• Type is readable at phone size without eating the demo.

If a cue fails, move that cue only. Do not restyle the whole job because one line was greedy.

CapCut will try to auto-reflow to bottom-center on a reimport. Descript will grow a box around a long line. Runway will "helpfully" center. Cut all three.

5. One human pass as a viewer who never heard the VO

Watch once on a phone with sound off, then once with sound on.

Check three traps:

• Captions that only work on a desktop timeline. If the phone fails, the burn fails.
• A line that names a button the hands are not pressing yet. Sync is a fact.
• A style so loud the demo looks like a lyric video.

Fix those with a move, a split, or a shorter line. Then burn. Do not open a "just in case" outline.

One pass. Ship the file. If a paragraph of caption still sits on the hands, you never added burned-in captions. Stop.

Reusable skill card

Trigger: You are about to click auto-captions or type "make the captions pop."

Input: The take, where the hands live, Y band, two-line max, plus the banned-phrase list.

Output: Burned-in captions in a locked band that misses the hands. No karaoke. No yellow brick. FAIL if a tap is covered.

Stop: After one mute-on phone pass. No second font. No lower-third in this pass. No VO cover-up.

Takeaways

• Lock a caption band away from the hands. Then burn once.
• Burned-in captions are speech. The tap is the proof. Do not sticker the finger.
• Mute the VO and watch the hands. Timeline zoom lies.
• Huge yellow type is a cover-up. Rewrite the VO if the words fail.