A lower third is not a caption track

A lower third is a name and title card. A caption track is speech, timed. They are not the same object, and they occupy the same band on a 9:16 frame. If you treat the super as “also captions,” you will cover the mouth, the username, and the line the mute viewer needed. Stack them on purpose, or drop one.

Two jobs, one band

A lower third is authored graphic: person, role, sometimes a company, held for two to five seconds while the plate moves. Nobody has to have said those words. The card is there so a viewer who walked in late knows who is talking. It is a title. It is not a transcript.

A caption track is what was said, aligned to the mouth, replaced every clause. Burned-in captions are that transcript as pixels. Auto-transcribe will never typeset “Jordan Hale, Floor Lead” unless Jordan said it. Running add_veed_captions on an interview and hoping the job title appears is a spend on speech you already had.

Overlays are the tool for the card. Captions are the tool for the talk. Mixing them in one preset is how you get a karaoke highlight on a surname the talent did not pronounce as a graphic, or a two-line speech block sitting on top of a name bar so neither can be read.

The confusion is geographical. Broadcast parked both in the lower third because 16:9 had room and the chyron language was already there. Vertical video parked platform chrome in that same third: username, caption field, CTA, shop tab. The “lower third” is now a contested strip, not a department. Safe area is the name of that fight.

This is not a contrast recipe for name supers on generated bokeh. Contrast still matters. The production error this page is for is identity: card versus track, then who wins the band.

They fight for the same safe area

On TikTok, Reels, and Shorts the bottom of the frame is already taken. Username and follow, the caption field, the right-rail buttons, on Shorts a photoreal AI overlay as of the 27 May 2026 YouTube Blog post. A broadcast-style name bar dropped on the true bottom is invisible. A caption block dropped on the true bottom is invisible. Both at once is two invisible objects and a covered face.

Decide the stack before you render:

• Speech is the product. Talking-head Reel, UGC ad, webinar clip. Captions own the band. If you still need a name, put the card high — top third, or a brief hold at the start — then get it out of the way so the transcript can sit where it can be read.
• Identity is the product. Expert lineup, panel, “who is this.” The card owns a short hold. Do not also run two-line captions through that hold. Either pause the transcript, or park captions higher for those seconds, or cut the card shorter than a sentence.
• Both are required. That is a layout, not a default. Two rows of speech plus a name bar is three lines. Three lines collide with chrome. You will drop a line. Pick which one before the client does it in comments.

Paid cuts add a legal row. That is a third graphic. The legal line is not a caption and it is not a lower third, but it wants the same pixels. BASIC two-line speech, legal as overlay, name card only if it fits above both. DYNAMIC word-pop plus a chyron plus a disclaimer is a poster, not an ad.

Do not steal the caption’s contrast treatment for the card and call it done. A name super that holds on a neon plate still fails if it sits on the username. Placement is the first QC. Contrast is the second.

Author the card; transcribe the talk

Workflow:

1. Lock the picture.
2. Place the lower third as an overlay with a start and end. Two to five seconds is enough. If the viewer still does not know who it is, the card is too small or too late, not too long.
3. Caption the speech on the rest of the file. Add captions to a video transcribes; it does not write the title. 165 language codes, 21 BASIC at 1×, 9 DYNAMIC at 2×.
4. Watch the overlap. If the card and a cue share a frame, one of them moves.

Speaker IDs in a caption track — JORDAN: — are not a lower third. They are a transcript convention for when the face is off screen or two voices share a mix. Use them in webinar clips and SDH. Do not use them as a substitute for a name super on a single talking-head who is already on camera. Redundant ID plus a card is two labels.

If nobody spoke, you do not have a caption job. You may still have a card. A silent product plate with a “Chef, 12 years” bar is overlay only. Running the captioner in any language is a listen on empty audio.

Caption Studio’s Headline group is a title row, not a transcript. Snap and Reveal are word-timed speech. Putting Headline on a full interview because it “looks like a lower third” will title every sentence. That is the kind mix again.

One band, one winner per beat

The mute phone pass is the judge. At each second, what is the one line the scroller must get?

• If it is the spoken claim, the caption is up and the card is gone or high.
• If it is the name, the card is up and the caption is held or parked.
• If it is the price or the legal line, that overlay is up and neither card nor karaoke is competing.

You can sequence those beats. You cannot composite all three at full size in the bottom 20 percent of a 9:16 and keep a face. Chrome will take its cut regardless of your design.

Reels and TikTok will not save a sidecar lower third. There is no sidecar for a chyron. Burn the card. Burn the captions. They are still two burns conceptually even when they land in one MP4.

The AI caption generator is the speech door. Overlays are the card door. Caption styles will preview type; they will not invent a job title. Keep the objects straight and the band gets simpler: one winner per beat, out of the chrome, readable on mute.

Stack the next card and track in Versely.

FAQ

Can auto captions generate a name and title card?

No. They transcribe speech. A job title nobody said is an overlay you write. Speaker IDs in a transcript are for off-screen or multi-speaker mixes, not a chyron.

Where should the lower third sit on TikTok or Reels?

Not on the true bottom. That band is username, caption field, buttons. Hold the card high, or hold it briefly at the start, then clear it for the speech captions. Check current chrome; safe area moves.

Can I keep captions running under a name super?

Only if both remain readable and out of platform UI. Two-line speech plus a card is already three rows. Pause or relocate the transcript for the hold, or drop the card sooner.