Animated captions: how to style captions people keep watching

What makes animated captions worth watching?

Captions on a short carry the words for everyone scrolling with the sound off, and give the eye something to follow. The caption styles that hold attention share a few habits:

Few words at once. Three on screen at a time is Sharpy's standard: readable at a glance, quick enough to keep pace.
The spoken word marked. A colour, a pop or a bolder weight on the word being said.
One word lifted. At most one keyword a phrase: the figure, outcome or feeling the line is about.
Contrast that holds. A dark enough picture behind white words, or an outline, shadow or box that makes one.
A clear place. Off the face, and out of the strips where TikTok, Reels and Shorts put their buttons.

A woman on a morning train watches a video on her phone without earphones, holding it close. — Many people watch with the sound off, so the captions carry the words and the rhythm of the voice.

Which caption styles does Sharpy have?

Twenty-two named styles, each built from its source's own numbers rather than a description of it, so Hormozi captions in Sharpy move the way Hormozi's do. Every style gives each word three states (unsaid, being said, already said), and lines are laid out at each word's widest state, so a word turning bold never shoves its neighbours.

Style	What the viewer sees
Hormozi	White capitals with one keyword in yellow
OpusClip's karaoke	Each word changes colour 75 ms before it is heard, and stays changed
Karaoke fill	Each word is wiped with colour as it is said
Submagic's Iman	Lowercase; the word being said turns bold without moving the others
TikTok	The whole line rises into place
MrBeast	Comic-book capitals
Kinetic slam	Words arrive four ways in turn: dropping, from the left, from the right, growing
Neon glow	Each word lights up and the picture darkens under the words

The shelf also holds Sharpy's house style, a green spoken-word style, pill karaoke, highlight, gradient, retro, explosive, classic, hype, vibrant, word-focus, minimal and a code-editor style.

How do you make animated captions in Sharpy?

In the window or by asking the agent; both work on the same edit.

Start from the words. In the window, one command makes captions from what was said in the edit. A video somebody else cut works too: Sharpy hears it with two transcription engines before burning anything in.
Add the script if you have one. Speech recognition misspells exactly the words a video is about. Sharpy fixes spellings from the script under one rule: the page may correct a word and may never supply one.
Choose a style. From the shelf, or through the agent's caption_video, which also sets the words on screen at once, their height and the highlight colour.
Set the type. The Text tab holds size, position, colour, capitals, typeface, weight, tracking, leading, alignment, outline width, how the words arrive, and five backings: none, pill, stroke, shadow and glow.
Adjust on the picture. Drag a caption to move it, drag a corner to size it, double-click to retype a word in place.
Render. Sharpy reads the captions back off the finished file and checks the highlight sits on the word being spoken.

Read back from finished renders, the highlight landed on the spoken word 63 times in 63 in one style and 132 in 134 in another, and a render that loses its highlight is refused.

How do keyword captions work?

A keyword caption lifts the one word in a phrase that carries the meaning, bigger, bolder or in another colour. The craft is consistent across the editors Sharpy studied: one meaning-carrying word a phrase, landed on the spoken word, held long enough to read, sized by importance, and never more than two lifted words on screen at once.

Ask the agent with suggest_keywords and Sharpy proposes each phrase's keyword with its evidence: the word the voice leaned on, and the craft's order of preference (a figure, an outcome, a feeling, the topic). A placed keyword stays anchored to its word after cuts, and Sharpy names what breaks the craft: an entrance more than half a frame from its word, a lifted connector, or a hold under five-sixths of a second, the shortest time Netflix's timed text guide allows a subtitle on screen.

Give it a reference video and it reads that editor's keyword language and applies it to your script.

Where should captions for reels sit?

Off faces. Sharpy finds the faces first; a line that would cross one goes below the lowest chin, or above the highest brow.
Clear of buttons. On a vertical piece, words in captions and graphics are held clear of the platform's own buttons. Turning a long video into shorts gives the strips to avoid.
Readable on any background. Each caption group is measured against the picture under it. Where white words would fall below 3:1 contrast, WCAG's minimum for large text, that group alone gets a heavier backing. On a reel shot against a light wall, captions read at 16.1:1, with none of 60 below the bar.
Never stacked. A render is blocked when two captions overlap in time and place.

How do you restyle captions or keep your own style?

To try another look, ask for restyle_captions: the edit's captions are laid again in another style, in the same place and height, without hearing anything again.

To make a look your own, start from any shelf style and keep it under your own name with your colours, typeface, words at a time, place and size; saved styles follow you through your account. Text looks go further: several outlines, a shadow, a box and a glow together, gradient fills, a shine across the letters, bevels and letters in depth. Sizes scale with the type, so a look holds at any size.

What should you check before you export?

The words. Read the list of words not taken from the script, and any the engines disagreed on.
The keywords. Each lifted word should be the point of its line.
The placement. Sharpy warns about text outside the title-safe area, or up too briefly to read.
The files. Tick the captions box in the export panel for SRT and WebVTT files too; they are written from the captions on the picture, so your corrections carry.

The next lesson covers subtitles, closed captions and SDH, and captions on your Mac covers every kind.

Frequently asked questions

How do you make animated captions?

Transcribe the video, time each word to the sound, then show a few words at a time with the spoken word marked by colour, weight or motion. Sharpy does all of it on your Mac from one of its 22 styles.

What caption style is best for TikTok?

One that reads at a glance: two to four words at a time in a heavy sans-serif, the spoken word marked, strong contrast, clear of the buttons on the right and the description at the bottom. Karaoke and Hormozi-style keyword captions are the common choices; match the style to the speaker's pace.

What font does Hormozi use for captions?

His team has not published it. Copies of the look usually use a heavy sans-serif in capitals, such as Montserrat Black, because the look comes from the capitals, the weight and the yellow keyword more than the typeface. In Sharpy you can set any typeface installed on your Mac.

Can I change the caption style after the edit is made?

Yes. Restyling lays the existing captions again in another style, in the same place and height, without transcribing again.

Is my video uploaded to make the captions?

No. Transcription, timing, styling and the render all happen on your Mac, with no limit on length.

Sources

Checked on 18 October 2026.