Snap captions scale and recolor each spoken word

Snap is kind=timed, effect=pop, Anton, accent #FF4785 on the base. Each word scales and recolors. Static subtitle groups cannot do this job.

Burn-in is not a generate. Snap is a caption preset you apply to speech that already exists on the file. It does not invent a line, it does not typeset a punch-out month, and it does not karaoke a silent product plate. If nobody spoke, you are on the wrong group.

The group is timed pop, not a subtitle card

The Snap group is five presets. All five are kind=timed. All five set effect=pop. All five sit at the bottom. All five use Anton — a display face, Heavy & Bold, a caption/overlay typeface, not an Ideogram poster and not a generate. The base is word_pop: white fill, black outline, accent #FF4785. Each spoken word scales up and jumps to that accent as it is said.

The other four variants change one field: the accent.

• Snap Gold — #FFD700
• Snap Cyan — #22D3EE
• Snap Lime — #A3E635
• Snap Mono — #FFFFFF (scale only, no hue jump)

Classic, Paper, Ink, Halo, Hype, Signature, and Headline are kind=subtitle. They burn a phrase. They do not scale a word. They do not recolor a word. Asking those groups to "feel like Snap" is a restyle of a different object. Pick Snap, or pick a subtitle group. Do not fake karaoke with a card.

Reveal is the other timed family: effect=typewriter, TikTok Sans, words appear and stay. That is a built line, not a pop. Fast talker, Snap. Building to a punchline, Reveal. Against a slow read, Snap looks frantic. Against a fast read, Reveal lags.

Speech in, pop out

Snap needs a transcript timed to the mouth. Add captions to a video is the transcribe-and-burn path: speech in, styled subtitles out. Apply the Snap preset on that pass. The words on screen are the words that were said. The pop is the effect. There is no second generate of type.

If the file is silent, Snap has nothing to scale. Overlay a CTA instead. If the file has music and no voice, Snap will not invent a hook. If the file has a voice and you wanted a headline nobody said, that is overlay, not Snap.

Anton is the face. It is not a model. Linking /fonts/anton is how you inspect the type. It does not render a poster. It does not letter an inspection tag. It sits in the caption stack.

Preview on your own take. A pop that reads on a demo clip can cover a face on a 9:16 close-up. Bottom is the group default. If the subject's mouth is in the lower third, the scale will hit the mouth. Crop or live with it. Do not ask Snap to become a top variant — this group does not ship one.

Timed overlays are a different clock

Add timed text overlays burns several authored lines, each with its own start and end. You write every string. You set every timestamp. Nothing is transcribed. Nothing pops a spoken word. That tool is add_timestamped_captions: a question at 0s, a stat at 3s, a punchline at 7s — copy you typed, not karaoke.

People confuse the two because both say "timed." Snap's kind=timed is word-level pop on a transcript. Timed overlays are beat-level type you authored. Use overlays when nobody said the line (a SKU, a month, a CTA). Use Snap when somebody said the line and you want each word to hit.

Do not:

• Run Snap on a silent product plate and hope the wrap becomes a script.
• Use timed overlays to fake a Snap pop. Overlays do not scale and recolor per word.
• Ask Veo to "caption in Snap." Veo will invent glyphs. Snap is burn-in after picture lock.

A generate of type is how you get a misspelled product name you cannot edit. Generate a clean plate. Transcribe or overlay. Then pick the group.

Lock the house style

Once Snap is the family, the variant is the brand. Accent is the only lever. Pink on the base, gold, cyan, lime, or mono. Switching to Hype on Thursday because the last Reel felt loud throws away the recognition the pop already earned. Write it down: Snap, Gold, bottom. Three words in a brief.

Mono is the quiet version: white to white, scale without a hue jump. Use it when the plate is already loud — a wrap, a racetrack, a neon lid — and you still want word timing. The effect size is the distance between resting fill and accent. Mono sets that distance to zero on color and keeps the scale.

Burn after crop. A bottom pop that sat clean on 16:9 sits in platform chrome after a 9:16 reframe. Picture lock, ratio lock, then Snap.

FAQ

Can a subtitle group pop each word if I pick a punchy font?

No. Classic, Paper, Ink, Halo, Hype, Signature, and Headline are kind=subtitle. They do not run effect=pop. Snap is the group that scales and recolors each spoken word.

Is Snap the same as timed text overlays?

No. Snap is a timed caption style on transcribed speech. Timed overlays are authored lines with start_sec / end_sec you typed. Overlays do not karaoke. Snap does not invent a line nobody said.

Does Anton generate the caption?

No. Anton is the overlay face Snap uses. Burn-in is an editor pass, not a text-to-image generate. Ideogram is a different row for posters and labels.

What if the take has no speech?

Do not use Snap. Overlay one authored line, or use timed overlays for beats you wrote. Snap with nothing spoken is a style waiting on a transcript that does not exist.