Caption Studio examples are the silent-watch test

The picker thumbnail is a still of a font. That is not how a caption works in a feed. Caption Studio's in-app examples — the same ten clips on this page — are one spoken line, ten VEED presets, burned in. Watch them with the sound off. That is the test. If the line does not land as a poster in the first second, the preset is wrong for the clip, no matter how pretty the type looks at rest.

Burn Versely captions in before you post is the strategy argument: platform auto-captions arrive late, in a typeface you do not own. Add captions to a video is the tool. This post is how to use the examples so you stop picking a style from a square.

One line, on purpose

The examples are not a catalogue of your content. They are a control. Same audio, same crop, same face, ten treatments. The only variable is the preset. That is the only way to judge furniture. If you audition styles on ten different clips, you are judging the footage, not the type.

The rail splits the way the studio does. The first six — Glass, Whisper, Fusion, Handwritten, Terminal, Glide 2 — are Dynamic: the animation is doing work. The last four — Simple, Corpo, Hustle, Vegas — are Basic: the type sits, the words change. Both families exist. The muted test is what tells you which job you actually have.

Mute first. Phone in hand, volume off, scroll the rail the way a stranger scrolls a feed. You are asking three questions, in order:

1. Can I read the first word before the face has done anything?
2. Does the motion help the line, or fight the mouth?
3. Would I recognise this account next Tuesday if the logo were cropped off?

If you cannot answer those from the example, you will not answer them from a thumbnail of Bebas Neue on a grey card.

What the muted test actually selects

A caption has three jobs. Only one of them is "the transcript." The examples exist so you can assign the job before you spend a credit on your own file.

Interrupt. Word-timed motion (the Dynamic row) is for the open: a number, a confession, a callout that has to arrive as an event. Glass and Whisper make the type feel like it is being said. Fusion and Glide 2 make it feel like a reveal. If your clip is a talking head whose first sentence is the hook, this is the row. A static subtitle on that open is a transcript of an interrupt, which is a contradiction.

Hold. Basic presets are for the body of a clip that already has a reason to stay — a demo, a founder read, a product in frame. Simple and Corpo keep the line readable while the picture does the work. Hustle and Vegas are louder Basics; they still do not animate per word. Loud at rest is not the same as loud in motion.

Furniture. After three episodes, a returning viewer should name the account from the caption before they name the product. Pick one treatment that survives mute, then stop. The caption styles catalogue is nine families of five — you are choosing a family, not a colour. Swapping Glass on Monday for Terminal on Thursday is a new show every time.

The mute test also kills a bad habit: picking the most animated preset because it looks expensive in the picker. Animation that lags the mouth or covers the product is a retention cut. You will only see that with the sound off.

How to run it on your own file

The examples tell you which family to lock. Your file tells you whether that family still works on this crop.

1. Shortlist three from the rail, muted. Not ten. Three. If you cannot throw seven away in a minute, you are shopping, not testing.
2. Burn those three onto the same cut. Add captions on the real audio, same start frame, same safe zone. Do not restyle between takes.
3. Watch the three exports muted, full screen, once each. No scrubbing. The first second is the whole test. If you need to rewind to read the line, the preset failed.
4. Lock the winner for the series. The next twelve episodes get that family. Colour variants inside a family are allowed. Crossing families is not.

Do this on a talking-head clip and on a product clip if you post both. A preset that clears a face can sit on the pack shot. The examples will not warn you about that; your crop will.

The transcript still has to be clean — a different pass — and none of this survives if the caption is a sidecar the platform restyles. The expensive mistake is treating the example rail as empty-state decoration. It is the cheapest A/B test in the studio: one line, ten styles, sound off. Use it that way, then generate.

FAQ

Why watch the examples muted if I already know the line?

Because the stranger does not have the sound on, and they do not know the line. Your memory of the sentence makes every preset look readable. Mute removes that cheat. If the first word is late, low-contrast, or sitting on the product, the clip has already lost the people who never unmute — which, on short-form, is most of them.

Should I pick a Dynamic preset for every talking-head clip?

No. Dynamic is for the open, or for a clip whose type is the hook. A 40-second founder read in Fusion is a music video of a paragraph. Basic on the body, Dynamic on the hook if you split them, one family across the account. The mute test will tell you when the animation is carrying the line and when it is just busy.

Can I mix two families in one video — a loud open and a quiet body?

You can. You should not, as a default. Two families in one file is two shows. If the open needs word-timed type, cut it as a hook and let the body use the same family at rest, or burn the hook as its own file. Recognition is furniture. Furniture that changes at second four is not furniture.

Do I still need the examples if I already locked a caption family?

Yes, once, on a new crop or a new format. A family that worked on a centred talking head can fail on a pack shot or a 16:9 cutdown. Re-run the mute test when the picture changes shape. Do not re-run it every Monday as an excuse to rebrand.