Chapter 23 — Key Takeaways
A one-page reference to screenshot before you shoot or cut a vertical. Specs change; the craft underneath doesn't.
The core idea
Vertical short-form is its own craft, not horizontal video cropped. The frame is a tall column, the opening is measured in two seconds, the captions carry the message for a muted audience, and retention — how much people actually watch — is the scoreboard that decides your reach. Everything from Chapters 1–22 still applies; it just bends around these constraints.
Definitions
| Term | Meaning |
|---|---|
| Vertical video | Video composed for a tall frame (usually 9:16) to fill a phone held upright, no black bars. |
| Safe zones | The region of the frame that stays visible and uncovered by platform interface and cropping. |
| The social hook | The first ~2 seconds engineered to stop a scrolling thumb and buy the next few. |
| Platform specs | The delivery requirements a platform expects: aspect ratio, resolution, fps, length, loudness, captions. |
| Retention / watch time | How much of a video people actually watch (avg. duration, % viewed, the retention curve). |
The vertical frame and safe zones
| Rule | Do this |
|---|---|
| The frame is a column | Compose vertically; get closer; favor a single subject. |
| Give the edges away | Assume bottom ~15–20% (username, caption, progress bar) and right ~10% (buttons) are covered. |
| Protect the center | Keep faces, key action, and text in the action-safe center column (~middle 80% width). |
| Frame a little loose | Center-safe framing survives every app's UI and any later re-crop. |
| Eyes on the upper third | Headroom still applies in vertical — turn on the phone's grid. |
| Keep verticals vertical | Use the level guide; a tilted upright frame reads as amateur instantly. |
The social hook — the checklist
Every strong social hook, checked in the first two seconds:
| ✓ | Channel / element | Question |
|---|---|---|
| ☐ | Visual hook | Is something striking or in motion in frame one? |
| ☐ | Verbal hook | Do the first words land on the payoff, not "hi / so today…"? |
| ☐ | Text hook | Is the promise/question on screen, legible muted? |
| ☐ | Open loop | Have I opened a question the viewer must stay to close? |
| ☐ | Front-loaded | Is my best moment first — even if it was shot last? |
| ☐ | No throat-clearing | Deleted the logo, greeting, and setup before the hook? |
Pick your hook type:
| Hook type | Opens the loop by… | Reach for it when… |
|---|---|---|
| Result / payoff-first | showing the end, then rewinding | you have an impressive transformation or result |
| Bold claim | daring the viewer to doubt you | you have a confident or mildly contrarian point |
| Direct question | naming what the viewer wonders | you can name your audience's exact pain |
| In medias res | starting mid-action, no setup | your footage has a strong moment of motion |
| Pattern interrupt | breaking the scroll's rhythm | you have something visually unexpected |
| Stakes / "wait for it" | promising a payoff worth the wait | there's real risk or a build worth teasing |
The hook is assembled, not filmed. Your best two seconds usually happen in the middle of the footage. Find them and move them to the front. When a video underperforms, re-order before you reshoot.
Captions & sound-off — the guide
Most feed video is watched muted. The video must work with the sound off.
| Caption rule | Why |
|---|---|
| Burn them in (open captions) | Guarantees the muted majority reads them; you control style/placement. |
| Large, high-contrast, with a stroke/plate | Readable on a phone at arm's length, over any background. |
| Center-safe band, not the bottom | The bottom strip is the platform's username/progress bar — they collide. |
| 1–6 words at a time, synced | Matches speech/beat; easy to read while watching. |
| Proofread the auto-captions | Auto-captions mangle names/jargon; an uncorrected caption is worse than none. ♿ |
| No strobing / hard flashes | Photosensitivity-safe motion. ♿ |
| Run the muted-comprehension test | Watch it silent — can a stranger follow it? If not, fix it. |
(Caption styling/animation → Chapter 34; burning/attaching captions on export → Chapter 36 §36.5; software button → Appendix E.)
Pace
| Do | Don't |
|---|---|
| Cut dead air — breaths, "um"s, pre-point pauses | Cut fast for its own sake (unmotivated = noise) |
| Keep something changing at all times | Leave static + silent stretches (exit ramps) |
| Cut on the sentence; ride a music/beat rhythm | Fade to black mid-video (an invitation to leave) |
| Keep a beat that lands a joke/reveal | Confuse "no wasted moment" with "as fast as possible" |
| Design a clean loop (last frame → first) | End on a dead-end sign-off |
Platform specs (a starting point — verify in Appendix H)
| Spec | Typical (at time of writing) |
|---|---|
| Aspect ratio | 9:16 vertical (short-form standard) |
| Resolution | 1080 × 1920 (upload native or higher) |
| Frame rate | 30 fps (24 or 60 also accepted) |
| Length | ~15–60 s sweet spot (maximums keep rising; short wins retention) |
| Loudness | ~ −14 LUFS (platform normalizes anyway) |
| Codec / file | H.264/H.265 .mp4, high bitrate (platform re-transcodes) |
| Captions | burned-in and/or uploaded caption file |
Aspect-ratio family: 16:9 (horizontal/long-form) · 9:16 (vertical short-form) · 4:5 (in-feed portrait — commands screen, survives the grid) · 1:1 (square, legacy).
Any spec you memorize will change. Confirm current values in Appendix H (Delivery & Codec Reference) and Appendix J (Resources & Communities); the deep delivery craft is Chapter 36 (§36.3).
Repurposing 16:9 → 9:16
| Method | When to use | Cost |
|---|---|---|
| Crop-and-reframe (keyframe to follow action) | There's room around the subject; most native look | Most work; auto-reframe is a rough first pass to fix |
| Stacked / letterbox | Cropping would lose info (wide demo, two-shot, text) | Keeps whole frame + adds caption space |
| Blur-fill | Last resort when neither works | Reads as a repurpose; wastes screen |
Best of all: shoot for the crop. Frame horizontal subjects centered and a little loose, essential action in a central column, so a vertical crop survives cleanly (Chapter 6, "compose for the crop"). You cannot cut a good vertical out of a bad horizontal.
Retention — reading the curve
| Curve shape | Diagnosis | Fix |
|---|---|---|
| Cliff at 0–3 s | Hook failed | Rebuild the first two seconds (don't touch the middle) |
| Gentle decline | Normal & healthy | Nothing |
| Mid-video sag/drop | A specific dull stretch | Find the timestamp; cut or tighten it |
| Late bump up | Rewatches / clean loop | Keep doing it |
Chase these, not likes: average % viewed / completion · shares · saves · rewatches/loops · follows. Views and likes are lagging vanity metrics.
Common mistakes → fixes
| Mistake | Fix |
|---|---|
| Composing edge-to-edge on the bare frame | Block out the bottom fifth / right tenth; keep the subject centered |
| Face on a side third (16:9 habit) in vertical | Recompose centered, eyes on the upper third |
| Opening with a logo / "hey guys" / setup | Cut it; lead on the payoff — the interesting part is frame one |
| No captions on a muted-first platform | Burn in captions; run the muted test |
| Trusting auto-captions raw | Proofread every word |
| "Crop to vertical later" with no plan | Compose for the crop, or use the stacked layout |
| Chasing max loudness / max length | Mix ~ −14 LUFS; make it as long as the idea needs, no longer |
| Steering by likes | Read the retention curve and shares/saves |
Throughlines, applied to short-form
- Story is the boss → retention is that principle with a scoreboard attached.
- Sound is half the picture → still true, but the picture+text must carry it muted; sound rewards the unmuted.
- You shoot for the edit → the hook is assembled in the edit; shoot centered so it crops.
- Motivate every choice → cut fast because the format punishes dead air, not for speed's sake.
- Fix it in pre, not in post → plan the reframe and shoot for the crop before you're stuck cropping badly.
This chapter's project step
Production Checkpoint (Projects 1–2): Cut a 15–30 s vertical teaser from Project 1 or 2 — reframed to 9:16 (subject safe-zone throughout), hooking in the first two seconds, captioned for muted viewing, exported to the typical vertical spec. Watch it back muted on your phone: would it stop your own thumb?
The one thing to remember
If you take one sentence into the feed: win the first two seconds honestly, make it legible with the sound off, and let retention — not likes — tell you what to fix.