Glass captions fail on busy plates

Glass captions fail on busy plates. The look is a transparent DYNAMIC burn: type sitting on the picture with almost no edge. Product image-to-video is specular labels, hands, chrome, and a pack that already has words. The caption disappears into the carton. Outline or a backdrop. Do not pick Glass because it is the default vibe.

Glass is a treatment, not a premium

On the VEED burn, Glass is one of 9 DYNAMIC presets, billed at 2×. Whisper, glide, fusion, terminal, handwritten, backdrop, and the rest of that tier are the siblings. 21 BASIC looks at 1× are the quiet set. Transcription is the same work. You are paying for how the glyphs move and how little furniture they carry.

Glass is popular because it looks like the picture is doing the work. On a dark talking-head with a clean wall, a thin transparent line can hold. On a product plate it does not. The plate is already type: a SKU, a claim, a window in the pouch, a lid with a word. Transparent caption type on printed type is two texts sharing pixels. Neither wins.

Caption Studio is a different list. Caption styles is 45 CSS presets, 9 groups × 5 variants. None of those groups is named Glass. Sheer variants in the card families (Paper, Ink) are the closest idea: a translucent card instead of a solid one. Sheer scores lower than the solid card it derives from. That trade is on purpose. It is also the same failure mode as Glass: you gave the plate permission to argue.

Do not pick Glass to “stay premium.” DYNAMIC is not better transcription. 2× is animation and a look. A wrong SKU in Glass is a pretty wrong SKU at double rate. BASIC 1× versus DYNAMIC 2× is the billing map. This page is the plate.

Product I2V is a hostile background

A pack shot is the product still that has to stay readable — label, shape, colour. Image-to-video treats that still as frame one. Motion around the pack — a hand, a pour, a turn — puts new highlights through the lower third every few frames. Pack shot is the object. The caption has to live with it, not on it.

Busy plates, specifically:

• Specular labels. Foil, gloss varnish, a wet bottle. A white Glass glyph on a white specular streak is gone.
• Already-lettered cartons. The SKU is in the frame. Caption type on the logo is a second headline.
• Hands and product chrome. Fingers cross the cue. Metal rims flash. The ratio is local, not a frame histogram.
• Lifestyle kitchens and windows. Generated UGC puts a sink, a tile, a blown-out window behind the talker. Glass on that wall is hope.

Generated plates make this worse on purpose: bokeh, neon, rain, a “cinematic orbit” that crawls the label. Hoping Veo 3.1 or Seedance will “keep the lower third dark” is not a caption strategy. The next take will not. The next b-roll cut will not.

WCAG contrast for burned-in type is the general rule: contrast is type on a moving plate, not a player setting. WCAG 2.2 AA is a claim about UI chrome. Burned-in still needs a ratio you can see on a phone. This page is the product case of that rule. Glass is the look that most often fails it.

Colour-only type fails the same way. Brand-accent cyan on a neon generate is not a contrast strategy. White on a pale pouch is not “clean.” It is invisible. Measure the worst frame behind the letters, not the brand hex on a grey picker.

Word-highlight on Glass is two problems. The live token may jump to a high-contrast accent while the rest of the line sits at a fail. The viewer still has to read the dim words. If only the pop is visible, you have a metronome on a busy plate, billed at 2×.

Outline or backdrop, not a darker generate

Treat the type as if the plate is hostile. Two treatments that actually hold:

• Stroke / outline. White fill, dark outline (or the inverse on a bright plate) manufactures an edge when the background will not. Too thin and compression eats it. Too thick and the word becomes a blob. Check at 1080×1920, not on a 4K timeline preview. Classic and Halo in Studio are this family of idea. BASIC “plain” only works if it carries a stroke and you still check the worst frame.
• Backdrop / scrim. A semi-opaque box or a local dim behind the cue. Paper and Ink are this idea: a card the plate cannot argue with. VEED’s Backdrop look is the DYNAMIC cousin — still 2×, still word-timed, but at least the card exists. You give up some of the picture. You keep the sentence.

A third treatment — picking a BASIC restrained preset and calling it accessible — only works if that preset carries a stroke or a card. Plain white with no outline is the most common miss on lifestyle footage, and it is cheaper than Glass, which is how teams ship a 1× miss instead of a 2× miss.

Do not fix a fail by going all-caps or adding emoji. Caps are taller and slower. Emoji is another glyph that needs its own contrast. Split the cue if CPS is tight. Change the treatment if the ratio is the problem. Those are different knobs.

Do not fix it by prompting the model to protect a lower third the captioner does not control. The generate will not remember. Put the edge on the type.

If the pack is the claim — a readable label in the last hold — the caption should move, not the crop. Covering the SKU to save Glass is the wrong object. Move type. Leave the label.

Check the worst frame, then burn

Order:

1. Lock the picture. Contrast QC on a cut you will trim is wasted.
2. Transcribe and fix names and numbers. A Glass miss on a wrong price is two errors.
3. Pick a look with an edge: outline or backdrop. BASIC at 1× is enough for product ads. DYNAMIC at 2× still needs that edge if you insist on motion.
4. Preview on the first seconds of this clip, then watch the whole export on mute, on a phone.
5. Pause on the worst frame of each cue — the specular flash, the hand crossing, the window. If the line goes, change treatment, not just colour.

The AI caption generator and add captions to a video will burn the preset you picked. They will not sample the plate for you. 165 language codes, 21 BASIC, 9 DYNAMIC — none of those is a contrast meter. Glass is a look. Busy plates are a condition. The look loses.

Closed-caption destinations still want a sidecar a player can restyle. That does not excuse a burned Reel with Glass on a white pouch. Short-form is the burn. The burn has to hold.

Glass captions fail on busy plates. Outline or a backdrop. Check the worst frame. Then burn.

Hold the next product burn against the plate in Versely.

FAQ

Is Glass always wrong?

No. It can hold on a dark, simple talking-head with no pack type in the band. Product I2V, foil, kitchens, and already-lettered cartons are the fail. If you cannot defend a ratio on the worst frame, pick an outline or a card.

Does a 2× DYNAMIC render fix contrast?

No. Billing is animation. Contrast is edge versus plate. Backdrop in the DYNAMIC set is the one look in that tier that brings a card. Glass does not.

Will a darker generate save Glass?

No. The next shot will not stay dark. Put the edge on the type. Do not prompt the model to protect a lower third the captioner does not control.