Learn the right vertical cover resolution, safe zones, text sizing, and an AI-assisted workflow for thumbnails that earn the swipe-stop on any phone screen.
A vertical video can be sharp, funny, and perfectly timed, and still lose its audience before a single word lands. In a swipe feed, the decision happens on one still frame — a cover image roughly the size of a postage stamp, competing against a dozen other covers while the viewer's thumb is already in motion.
That frame appears in more places than most creators assume. It shows up on your channel's vertical shelf, in search results, in subscriber feeds, in the related rail after someone finishes a different video, and as the preview card when a link is pasted into a chat. Some surfaces show a full 9:16 crop. Others show something closer to a square. A few crop a fragment and lay interface elements directly on top of it.
This guide gives you a working specification for that still frame plus a repeatable production pipeline, so you never rebuild a template from scratch at publish time. We will work through resolution and aspect ratio, safe zones, composition for a narrow canvas, typography that survives downscaling, a step-by-step export routine, where AI generation genuinely helps, the mistakes that quietly flatten reach, and a QA pass you can finish in three minutes.
The Numbers Behind a Vertical Cover
Why 1080 × 1920 Is the Working Standard
1080 × 1920 at 9:16 is the canvas. Design at that size and export at that size. Anything smaller gets upscaled by the player and your type goes soft and slightly fuzzy around the edges, which reads as amateur even when the underlying composition is strong. Anything larger gets downscaled and you lose the crispness you spent render time producing, sometimes with shimmer on fine detail.
Long-form thumbnails use 1280 × 720, and that difference matters more than people expect. Do not build a tall cover by stretching a landscape frame. The composition logic is different, and a stretched crop usually slices a face in half or leaves dead bands of empty background at the top and bottom that make the cover look unfinished.
Building a Master That Survives Every Crop
If you want a single asset that behaves well everywhere, design a 1080 × 1920 master but confine every essential element — face, headline, wordmark — to the centered 1080 × 1080 square. Then, when a surface displays a square or 4:5 crop, your subject and headline are still intact. Everything outside that square is texture, gradient, or background color, and losing it costs you nothing.
In practice you should export three files from one design: 9:16 (1080 × 1920), 4:5 (1080 × 1350), and 1:1 (1080 × 1080). One master, three crops, no redesign work. Keep the master project file so a later refresh takes minutes instead of an hour.
File Formats and Compression
PNG is the safest still format when you have crisp type or flat color blocks, because it avoids ringing artifacts around high-contrast edges. High-quality JPEG is fine for photographic frames and produces far smaller files. Keep the PNG as your archive and publish the JPEG when a file size ceiling matters. Platform thumbnail rules for long-form video — under 2 MB, at least 640 pixels wide, JPG, PNG, or GIF — are a reasonable hygiene floor even when your publishing path for a short differs.
One habit saves more quality than any setting: never re-export a frame you already pulled from a compressed delivery file. Go back to the source timeline or the original render. Otherwise you stack compression on compression, and edge detail dissolves into a haze that no amount of sharpening fixes.
Safe Zones: Where the Interface Eats Your Art
Think of the vertical frame as having four layers of obstruction that you cannot control and cannot preview perfectly.
| Region | Typical obstruction | Keep out |
|---|---|---|
| Top 8–10% | search bar, back arrow, status overlays on some devices | headline, wordmark, faces |
| Right 12–15% | like, comment, share, and remix rail during playback | faces, text, key detail |
| Bottom 20–25% | channel name, caption snippet, sound label, action buttons | anything you want read |
| All edges ~5% | rounded-corner masks, unpredictable chrome | critical detail of any kind |
A practical rule: define an inner safe rectangle of roughly 960 × 1440 anchored slightly above center. Headline, logo, and face live inside it. Everything outside is texture, background, or color field.
Why the Bottom Quarter Is the Danger Zone
The most-viewed surface is the player itself, and the player covers the lower portion of the frame with a caption block and action buttons. A punchline headline tucked into the lower quarter is simply invisible on the surface where most people will see your video. Move it into the middle band or just below the top edge. This single change fixes more underperforming covers than any other adjustment.
Testing Safe Zones on Real Devices
Check your design on at least one small phone and one large one. Emulator previews are convenient but they do not always reflect where chrome lands, especially when a device has a taller status bar or a different navigation gesture area. A quick habit: screenshot your published post after it goes live, then drop that screenshot into your design file as a locked reference layer for the next cover in the series.
Vertical Composition Is Not Cropped Landscape
Landscape images read left to right. Vertical images read top to bottom. That single difference changes how you build a hook. A left/right split that works in 16:9 becomes two cramped columns in 9:16, both of which feel squeezed. Stack instead: a visual anchor in the upper third, a headline band in the middle, supporting texture below it.
The rules that hold up across most successful covers:
- Place the subject's eyes in the upper 35–45% of the frame, because the top half of a tall image registers first in a feed grid.
- Give faces room. A tight crop that looks fine on a desktop monitor feels claustrophobic at 180 pixels wide.
- Reserve one full-height column as negative space for type instead of adding another heavy overlay box.
- Aim for three visual weights and no more: one dominant element, one supporting element, one texture layer.
The Three-Weight Rule in Practice
If your Short is about a surprising reaction, the cover is the reaction. If it is about a transformation, the cover is the before-and-after split. Adding a headline, a logo, an arrow, and a second face pushes the cover into visual static, and static does not stop a thumb. Ask yourself which single element a viewer would remember after one glance, then delete or fade anything that competes with it.
Depth Cues That Suggest Motion
Stills that imply movement hold attention longer, because the brain predicts continuation. Slight motion blur on a limb, a trailing edge of fabric, drifting particles, or a shallow depth-of-field falloff all signal that something is happening. If your source footage is flat, you can build depth in post with a subtle vignette, a gradient scrim, or a blurred duplicate layer behind the subject. None of these tricks add information, but they add atmosphere, and atmosphere is what makes a still feel like a moment instead of a frame grab.
Typography: Cutting Words Until It Hurts
Text on a vertical cover is a scalpel, not a sledgehammer. Treat every word as something that must earn its space at thumbnail size.
Keep it to two to four words. A five-word headline can work if the words are short and the type is condensed, but anything longer turns into gray mush at preview size.
Size for the smallest preview. On a 1920-pixel-tall canvas, a cap height around 110–150 pixels is a reasonable starting point for a headline. Shrink your design to 20% and read it. If you cannot, it is too small, no matter how elegant it looks at full size.
Use one typeface family. A heavy condensed sans for the headline plus a regular weight for a supporting line is plenty. Script faces and thin display fonts collapse into illegible strands once downscaled.
| Element | Starting size on a 1080 × 1920 canvas |
|---|---|
| Headline cap height | 110–150 px |
| Supporting line cap height | 55–75 px |
| Wordmark height | 50–70 px |
| Longest headline line width | 860 px maximum |
Build contrast deliberately. Add a soft dark scrim, a solid color bar, or a subtle drop shadow rather than hoping the background happens to be dark where the type lands. Aim for a contrast ratio of at least 4.5:1 between the text and the pixels directly behind it.
Do not repeat the video title. The title usually appears next to the cover, so duplicating it spends your best real estate on information the viewer already has in front of them.
Handling Long Words and Non-Latin Scripts
Long compound words in German or Polish, and scripts that do not use spaces, need different line-breaking logic. Set the headline block to a fixed width, allow a maximum of two lines, and cut words until it fits. If a two-line headline runs past the safe rectangle, the fix is fewer words, not smaller type. Smaller type at this scale is the same as no type at all.
A Repeatable Workflow From Timeline to Published Cover
Once the template exists, this sequence takes under ten minutes.
1. Shortlist three candidate frames. Scrub the timeline and grab the frames with the strongest expression, gesture, or composition. Do not settle on the first frame that looks acceptable.
2. Normalize the grade across candidates. Apply the same color treatment to each so your choice is based on composition, not on which frame happened to be brighter.
3. Export at full resolution. Pull each still from the source timeline at 1080 × 1920 or larger. Never screenshot a preview window.
4. Composite inside a locked template. Safe-zone guides, headline block, wordmark, and optional accent shape sit on fixed layers. You only swap the underlying photograph.
5. Export a PNG master and a compressed JPEG. Keep both. The PNG is your archive; the JPEG is what you publish when size matters.
6. QA on an actual phone. View at 100%, then at roughly thumbnail size, then beside screenshots of your last nine covers. This step catches problems a large monitor hides.
7. Name and archive consistently. A convention like series_ep12_cover_v3.png saves hours when you rebuild a series look six months later.
Building the Template Once
Spend twenty minutes on the initial setup: layer order, locked guides, text styles, a saved color palette, and export presets for the three crop sizes. That investment removes almost all repetitive decisions from the weekly routine.
When You Publish Daily
Batch your covers. Design a week of them in one sitting with the type layer identical across the set and only the photograph changing. Consistency is easier to maintain in a batch, and you will notice composition errors faster when nine covers sit side by side.
Where AI Generation Earns Its Place
AI is most useful at three points in this pipeline and least useful at exactly one.
Generating Backgrounds and Set Pieces
When your footage is plain — a talking-head clip against a neutral wall, for instance — you can generate a rich background or a stylized set piece and composite your subject over it. A dedicated AI image generator is the right tool for that job: you get consistent, prompt-controlled stills you can refine across an entire series without reshooting anything.
Baking Composition Into the Prompt
When the video itself is generated rather than filmed, the cover frame becomes a design decision you can specify up front. Vertical-friendly prompts describe subject placement, headroom, negative space reserved for a headline, and lighting direction. Browsing a prompt library is a fast way to learn which phrasings produce compositionally usable frames instead of centered, symmetrical images that leave no room for type.
If your clip comes out of an AI video generator, render extra takes and treat the last second of each as a candidate cover. Generated motion often peaks visually just before it settles, which is exactly the instant that reads well as a still.
Keeping a Series Coherent
Style consistency across episodes is where generation earns its keep. Lock a prompt skeleton — lens, palette, lighting, framing — and change only the subject line. You end up with a family of covers that reads as one series, which is the recognition cue that makes viewers click episode five because they liked episode four. Starting from a fixed template also means the type layer is never rebuilt by hand.
The One Place AI Still Underwhelms
Fine typography and precise logo placement. Generated images struggle with clean letterforms and exact spacing, and the failures are obvious at small sizes. Generate the photographic layer, then set your type in a proper editor with real font rendering and real kerning controls.
One Master, Several Vertical Feeds
Different feeds crop differently, offer different cover-selection behavior, and place different chrome over the frame. Rather than chasing each platform separately, work from a single disciplined master.
- Export 9:16, 4:5, and 1:1 from the same design so you can match whatever placement you are publishing to.
- Keep the visual anchor in the upper half, because placements that crop from the center will keep the middle and discard the edges.
- Check the link preview card separately. It often centers on the middle of the frame and can crop out a headline that sat near the top.
- Decide the first frame deliberately, since some placements fall back to it when no custom cover is applied.
The First Frame as Insurance
If a custom cover is not available on every surface, your opening frame becomes the cover by default. Design frame one with the same discipline as a cover: face in the upper half, no critical detail in the bottom quarter, and a clean region where type could be added in post. When a video opens on a well-composed frame rather than a mid-motion blur, the fallback case stops being a liability.
QA Checklist and the Mistakes That Cost Reach
The most common failures are consistent, boring, and completely fixable:
- Text sitting in the bottom quarter, where player chrome covers it on the most-viewed surface.
- Overcrowded composition: three subjects plus a headline plus a logo equals noise.
- Low-contrast type, such as white on a bright sky, which reads as a coin flip.
- A reused landscape crop, leaving empty bands at the top and bottom.
- A face sliced by the crop mask when the square or 4:5 version is displayed.
- Oversharpened edges that shimmer when the image is downscaled.
- A wordmark colliding with the right-hand action rail.
- Inconsistent series branding, so viewers cannot tell your episodes apart in a feed.
Run a six-point check before every upload: safe zone respected, headline readable at 20%, contrast ratio at least 4.5:1, face unobstructed, brand mark present and aligned, and file under your size threshold.
Testing Without Fooling Yourself
Change one variable at a time — headline wording, accent color, or subject framing — and compare across a reasonable sample of uploads rather than a single post. Signals worth watching include how quickly viewers swipe away in the first two seconds, average view duration, and how often a video travels beyond your subscribers. A cover that lifts watch time on the first swipe is doing its job, even if it looks less polished than the one you liked more.
FAQ
What resolution should a vertical cover be? 1080 × 1920 at 9:16. Design and export at that size so nothing gets resampled by the player.
Why does my cover look blurry after publishing? Usually one of three causes: it was exported smaller than 1080 pixels wide, it was captured from a preview window instead of the timeline, or it is a re-export of an already-compressed frame. Fix all three and the softness disappears.
How much text is too much? If you cannot read the headline at 20% zoom, it is too much. Two to four words is a reliable range; two lines is the practical maximum.
Should vertical covers match my long-form thumbnails? Share the typeface, accent color, and wordmark, but do not share the layout. Landscape and vertical compositions need different structures, and a stretched landscape crop almost always reads as an afterthought.
Do AI-generated covers hurt performance? Only when they look generic. Audience-tested, prompt-controlled stills with deliberate typography perform well; obvious default outputs with a centered subject and no negative space do not.
What is the single biggest mistake? Placing important content in the bottom quarter of the frame, where player interface elements cover it on the surface that matters most.
Do I need a custom cover at all if the first frame is strong? A custom cover gives you a designed hook and a consistent series look. A strong first frame is insurance for placements that ignore custom covers. Ideally you have both, built from the same composition logic.
How often should I refresh the template? When performance plateaus or when the style starts to feel dated. A couple of times a year is usually enough, and each refresh should change one or two elements rather than everything at once.
Turn Cinematic Ideas Into Motion With Orelon
The best cover frame is not something you hunt for after the edit is finished. It is the shot you designed on purpose: with headroom for a face, a clean column for a headline, and lighting that still reads when the image is 180 pixels wide. Orelon is an AI video generator for cinematic ideas in motion, built so you can compose vertical scenes for a small screen from the first prompt instead of fixing the framing in post.
Start from the Orelon homepage to see how the workflow fits together, then build your first vertical scene with composition decisions already baked in. When the cover and the opening frame come from the same intentional design, the swipe-stop stops being a gamble.