开场中的画面、口播与屏幕文字

Every short-form hook runs three things in parallel during the first 1 to 2 seconds: a visual layer (what the eye sees), a verbal layer (what the voice says), and a text layer (what the on-screen caption asserts). The strongest hooks have all three pointing at the same idea. A clever spoken line on a flat first frame is half a hook.

Some people add a fourth: the audio layer (the actual soundscape, a beat drop, a record scratch, an abrupt silence). When it works, it acts as the timing cue that locks the other three together.

Why all three matter

• A large share of short-form views happen on mute. If the verbal hook carries the whole promise and there's no caption, most viewers see a silent face and swipe.
• The eye registers a visual before the brain finishes parsing a spoken sentence. Pre-attentive processing fires inside the first half-second. The first frame has to do work before the first syllable lands.
• Captions tend to lift completion on short-form by a measurable amount versus the same video without them. Treat captions as required, not optional polish.

The visual layer

What the eye gets in the first frame.

• Show the payoff or the conflict. A finished result, a strong emotion, a side-by-side, the after of a before-after. Not a static face waiting to begin.
• Compose for paused-player legibility. The first frame should read the moment a viewer's thumb pauses on the feed. If it needs the next two seconds to make sense, it's not a hook frame.
• Use motion as a pattern interrupt. Movement, a sudden zoom, a jump cut, a hand entering frame. Anything that breaks the rhythm of the surrounding feed.
• Click-to-unpause framing. Compose the opening frame as if it has just been paused mid-action. The viewer's instinct is to play to see what happens next.
• No logo, no countdown, no fade-in. Every frame of throat- clearing is a frame of swipe.

If the only difference between your first frame and the next creator's is the face, your visual hook isn't doing work yet.

The verbal layer

What the voice says.

• Five to ten words. Short enough to land before second 2. The spoken line is also indexed for search on most platforms, so it doubles as discovery copy.
• Lead with the payoff or the stake. "Your $300 coffee setup is why it tastes worse" lands. "Hey everyone, today I want to talk about coffee" doesn't.
• One claim, said once. No setup, no autobiography, no "so basically." If the verbal hook needs a sentence of warm-up, the warm-up was the real hook. Start there.
• Use power verbs. "Burned" not "spent." "Cracked" not "figured out." Verbs imply motion and stakes; weak verbs imply a meeting.
• Match the energy of the visual. A high-stakes line on a slow static shot reads as disjointed. The voice has to live where the eye is.

The on-screen text layer

What the caption asserts in the first frame.

• Three to seven words. Long captions don't read in the first second.
• Sharpens, doesn't repeat. If the voice says "I spent four hundred dollars on coffee gear," the text doesn't repeat the sentence. It says "$400 burned." It compresses, escalates, or states the stake.
• One idea, big and bold. High contrast against the background. Center or upper-third placement so the platform UI doesn't sit on top of it.
• No emojis as content. They eat character budget and read as filler.
• Front-load the most interesting word. If only the first two words read, those two words should make the swipe pause.

A good test: cover the text and ask whether the hook still works muted with just the visual. Then cover the visual and ask whether the text alone earns a pause. Both should pass.

The audio layer (Kallaway's fourth)

What the soundscape does.

• A beat drop on the first cut. A record scratch on the contrarian line. A sudden cut to silence on the reveal. The audio is the timing cue that locks the other three layers together.
• Trending sound, when it fits, signals freshness to the algorithm and to the viewer at the same time. When it doesn't fit, it just reads as noise.
• Original audio is the right call for talking-head and demo content where the voice has to land.

Designing all three together

Practical order for drafting:

1. Write the verbal line first. It's the easiest to test by reading out loud. Cut until it's under 2 seconds.
2. Design the visual to land before the voice does. The first frame either shows the payoff, the conflict, or the stake. If the voice has to explain what we're looking at, the visual isn't pulling weight.
3. Write the text last, as a sharpener. Compress the verbal line. Don't repeat it.
4. Pressure-test alignment. If a viewer watches muted, do they get the hook? If they look away, does the voice still pull them back? If they're glancing while scrolling, does the text alone make them pause?

A misaligned three-layer hook reads as confused. An aligned one reads as inevitable. The difference is almost never the cleverness of the line. It's whether the layers agree.

See assets/three-layer-worksheet.md for a side-by-side sheet to fill in before shooting.