Captions on faceless product demos

On a faceless product demo, type is the presenter. There is no mouth to watch. The mute scroller reads the caption or they bounce. BASIC two-line is the demo body. DYNAMIC karaoke is only for a spoken hook you can still finish. Do not treat a pack-only clip like a talking-head Reel.

There is no face to carry the line

Talking-head UGC has a person. Even muted, a face holds. A faceless demo has hands, a carton, a screen recording, a tabletop. UGC that is only product still needs a claim in the picture. Burned-in speech is that claim when someone spoke. Authored overlay is that claim when nobody did.

Two different faceless jobs get collapsed:

• Spoken demo. A VO or native-audio bake explains the product while the plate is hands-on. Transcribe and burn. The caption is the presenter.
• Silent plate. Music, whoosh, no words. ASR will not invent the demo. Write the overlay. Running add_veed_captions on a silent pour is a spend on empty audio.

Native-audio generates still need the burn. A Seedance or Veo take that talks over a product is diegetic speech in a mute feed. Caption it. Silent turbo rows need a VO first. Either way the face is absent, so the type has to do the job a chin would have done: tell the scroller what this is.

Faceless is not a licence to fill the frame with paragraphs. Two lines. Finishable. Off the SKU. The pack is the talent. Covering the label to make room for the “presenter” is how you fire the talent.

Kapwing’s 2026 UGC round-up (with Billo, archive.com, creator.co in the same set) is why product-native video is in the account: on the order of 4× CTR and about 50% lower CPC versus non-UGC in that research stack. None of those figures is a caption study. They are why a faceless demo still has to work as a muted ad. Type is how it works.

BASIC is the demo; DYNAMIC is a hook you actually said

add_veed_captions is one engine. 21 BASIC looks at 1×: line-level, the whole cue arrives together. 9 DYNAMIC looks at 2×: word-level animation. 165 language codes on both. Accuracy does not improve at 2×.

A product demo is usually a body, not a punchline. The viewer has to follow a sequence: open, pour, swipe, result. Line-level BASIC holds that sequence. Word-pop on every instruction is a strobe over the hands. Courses and how-tos already learned this. Faceless demos are short how-tos. Treat them that way.

Use BASIC when:

• the clip is a procedure (more than one beat)
• a price, SKU, or legal line already occupies a band
• the plate is a busy pack (Glass will die; a quiet outline will not)
• you are captioning a month of SKUs and the look should not double the meter

Use DYNAMIC when the first two seconds are a spoken phrase — “this stain is gone in one pass” — short enough to read, and the rest of the demo drops to a static block. Karaoke as a hook, not as a twelve-step highlighter. Snap and Reveal in caption styles are the Studio version of that hook clock. Pin one family for the series. Do not Snap the open and Paper the body and Hype the CTA. That is three shows.

If nobody spoke the hook, karaoke cannot typeset it. “WAIT FOR IT” over a silent pour is an overlay. The AI caption generator transcribes. It does not write the demo script.

CPS is tighter without a face. There is no lip-reading assist. If you cannot finish the line on mute at phone size, split or cut. All-caps “demo energy” is slower and taller. Skip it. Sentence case, outline or a card, two lines.

Place for the pack, not for a mouth that is not there

Talking-head defaults assume a chin in the lower third. Faceless defaults should assume a label there.

Keep-outs:

• Bottom chrome: username, caption field, CTA, Shorts AI overlay on photoreal generates as of 27 May 2026
• The pack shot: letters on the carton
• Hands crossing the mid-frame

The hole is often mid-upper. Top variants exist in several Studio families. Use them. Top chrome still exists; do not sit in the status bar. Safe area is a snapshot. Recheck in the app.

Contrast on product plates is hostile: specular, already-lettered boxes, kitchens. Glass fails. Outline or a backdrop. BASIC plain with no stroke is a miss on a white pouch. The presenter has to be readable. A transparent vibe is not a presenter.

Numbers must match the PDP. Faceless demos are often shoppable. A captioned “$19” on a $21 page is the ad. Search the transcript for price, percent, SKU. ASR will miss them.

B-roll cutaways inside the demo need a re-time. A two-second instruction cannot sit on a half-second insert. Caption after the edit.

One presenter look, then burn

Write the series spec: family or VEED id, BASIC unless a named hook, position off the pack. Every SKU in the line uses it. Changing family per product reads as a different store.

Order:

1. Lock the demo cut. Hands, pack, result.
2. Confirm there is speech. If not, author overlays and skip the captioner.
3. Transcribe. Fix names and numbers to the PDP.
4. Place two lines in the hole that is not chrome and not label.
5. Burn BASIC. DYNAMIC only on a spoken hook you previewed on this plate.
6. Mute phone pass. If the presenter (the type) disappears on a specular frame, change treatment. If it covers the SKU, move it.

Add captions to a video will burn the file you give it. Give it the faceless cut, not a talking-head preset you liked on a different series. 45 CSS presets, 9 × 5. Pick one group. The presenter should look like the same person every Tuesday.

A shoppable faceless cut is still an ad. The PDP the swipe opens has to match the digits in the presenter type. If the VO said last week’s price, you recut or you overlay the live number — you do not let ASR publish a stale offer because there was no face to “look trustworthy.” Trust on this format is the pack plus the words.

Faceless is not silent. Mute autoplay still needs words. On this format those words are the talent.

Burn the next demo presenter in Versely.

FAQ

Do silent product plates need VEED captions?

No. There is nothing to transcribe. Write overlays. Running the captioner on music-only audio is a spend on empty. If you add a VO, caption after the VO.

Should every faceless demo use karaoke because there is no face?

No. No face is a reason to be more readable, not more animated. BASIC two-line is the presenter. DYNAMIC is a short spoken hook, if you have one.

Can I put the steps in three lines so the demo is complete?

Height is scarce. Three lines hit chrome or the pack. Split into two-line cues over time. Keep the type large enough to be the presenter.