๐Ÿ  Hub ๐ŸŽฏ Optimizer ๐Ÿ“… Planner โœ“ Tasks ๐ŸŒฑ Growth playbook

Thumbnail Checker ๐Ÿ™

Drop in a thumbnail and I'll tell you if it'll actually get the click. Built for YouTube long-form and Twitch stream previews. โœจ

1

Drop your thumbnail PNG, JPG, or WebP

Minimum 1280ร—720, ideal 1920ร—1080 (16:9, under 2MB โ€” big screens are a growing share of views). I'll grade it for click-through readability and show you what it actually looks like at every size people will see it.

Your thumbnail
Loading modelsโ€ฆ
First time runs download ~240MB of model files. Cached after that.
๐Ÿ“š

Thumbnail wisdom what VidIQ + top creators have proven works

These are the rules the scoring rubric is built on. Each one comes from observed CTR data on hundreds of thousands of YouTube + Shorts thumbnails. Click any card to expand the details.

๐Ÿ˜ฑFaces with emotion+20โ€“30% CTR

Thumbnails with a clear face โ€” especially showing shock, joy, disgust, awe, or fear โ€” outperform face-less thumbnails by ~20โ€“30%. Brains are wired to lock onto faces in the periphery; emotion adds urgency. The catch: a neutral face performs worse than no face at all โ€” emotion is the whole trick, not the face itself.

  • Do: Crop the face large (1/4 to 1/3 of the thumbnail). Eyes looking AT the camera or AT the subject of interest. Exaggerated-but-authentic expressions win.
  • Don't: Tiny face in corner, neutral expression, fake open-mouth shock, closed-mouth smiles, sunglasses, fully obscured face.
  • VTubers: Your model counts โ€” use the most expressive frame from your stream. Auto face detection misses 2D / illustrated models, so the character checklist in Step 2 lets you mark what's actually in your frame.
๐Ÿ”คText: 2โ€“4 words MAXreadable at 120px

YouTube mobile thumbnails render at ~120ร—68px. If your text isn't readable at that size, it's noise. The sweet spot is 2โ€“4 huge words.

  • Do: "I QUIT", "FACE REVEAL", "$10K BUILD", "THIS IS WHY".
  • Don't: Full sentences, watermark text, channel name across the thumbnail.
  • Font: Bold sans-serif (Impact, Anton, Bebas Neue). Outline + drop shadow so it survives any background.
๐ŸŒˆHigh contrast + saturationscroll-stopping

The mobile feed is a stack of muddy, low-contrast images. A saturated, high-contrast thumbnail interrupts the scroll. Black-to-white luminance variance > 0.18 is the magic number.

  • Do: Bright accent colors (hot pink, neon green, electric blue) against dark/muted backgrounds.
  • Don't: All pastels. All dark. All one hue. Grey-on-grey.
  • Tip: Use complementary colors โ€” orange + blue, purple + yellow, red + cyan. They vibrate against each other.
๐ŸŽฏOne clear focal pointrule of thirds

The eye should know IMMEDIATELY where to look. One subject, dominantly placed. Anything else is noise.

  • Do: Place the subject at a rule-of-thirds intersection, not dead center (unless the composition demands it).
  • Don't: 5 different elements all fighting for attention. Crowded backgrounds. Multiple focal points.
  • Negative space: Keep 30โ€“40% of the frame calm โ€” clutter drowns the focal point at feed size. The score's "Negative space" check measures this.
  • Test: Squint at the thumbnail from 6 feet away. Can you tell what it's about in 0.5 seconds? If not, simplify.
๐Ÿ“ฑMobile-first: 168ร—94px test70%+ of views

70%+ of YouTube views come from mobile. Most viewers see your thumbnail at roughly 168ร—94px in the feed โ€” and they decide whether to click in under one second. If it doesn't pop at that size, it doesn't matter how good the desktop version looks.

  • Do: Test by scrolling past it on your own phone. Does it stop you?
  • Don't: Pack details only visible at 1280px. The duration badge (bottom right) and progress bar cover the bottom 10% โ€” keep critical text out of there.
  • Tip: The mobile preview above this page shows you exactly how it'll render.
๐ŸŽจPattern interruptscuriosity gap

Best-performing thumbnails break a visual or contextual expectation. Something is "off" in a way that demands explanation.

  • Visual: Subject in an unexpected location, impossible scale, dramatic lighting, oversized prop.
  • Contextual: A face that doesn't match what the title implies. A title that contradicts the image.
  • Avoid clichรฉs that have stopped working: Generic red arrows. Open-mouth shock that looks fake. The "circled face" trope without purpose.
๐ŸŸฆEdge safety: stay 10% insidevertical-safe

YouTube and other platforms crop thumbnails differently across surfaces (especially Shorts vs long-form vs Twitch). Keep your face + text in the inner 80% of the canvas โ€” the outer 10% on every side may get cropped.

  • Bottom: duration badge eats the lower-right corner.
  • Top: some platforms overlay a "new" badge or live indicator.
  • Tip: If your subject's eyes are off-screen on the YT mockup above, you've got an edge-safety problem.
โš–๏ธMatch the title's promiseretention > clicks

A clickbait thumbnail with a mismatched video = high CTR, terrible retention = the algorithm punishes you. The thumbnail and the first 8 seconds of the video must deliver on the same promise.

  • Do: Use thumbnails from actual scenes IN the video. The viewer should think "yes, that's the moment I saw."
  • Don't: Bait with something that never appears. Use scenes from a totally different video.
  • Test: If a stranger saw your thumbnail and title, would they accurately predict what the video is about?
๐ŸงฉTitle + thumbnail = one packageinformation split

The thumbnail and title are your only two persuasion surfaces, and they must not repeat each other. Thumbnail = emotion + visual proof. Title = keyword + promise. Together they open ONE curiosity gap the viewer needs closed.

  • Never duplicate text: if the title says it, the thumbnail shouldn't. Paste your title in Step 1 and the synergy check flags every repeated word.
  • Package-first: top creators judge an idea by whether it survives as a title + thumbnail BEFORE filming. If you can't package it, the idea is weak no matter the execution.
  • Auto-play era: home-feed videos now auto-play, so the first 5 seconds of the video are an extension of the thumbnail โ€” open on the visual question the thumbnail asked.
  • Re-package before you give up: a thumbnail swap or a tighter title has flipped daily views by large multiples on otherwise-dead videos. 90% of top-performing videos use custom thumbnails.
๐ŸงชA/B test like Studio does1 variable per variant

YouTube Studio's "Test & compare" runs up to 3 thumbnails for up to 2 weeks and picks the winner by watch-time share, not raw CTR โ€” a lower-CTR thumbnail can win if it attracts viewers who stay.

  • Change exactly ONE variable per variant (color, text, expression, OR composition) โ€” change two and you learn nothing from the result. The ๐Ÿ“Œ Pin-for-A/B button on this page works the same way: pin, change one thing, re-upload.
  • Small margins are real: a 0.5% CTR difference is significant at scale, and documented tests have moved CTR 37โ€“110%.
  • Know your baseline: average CTR by niche โ€” Gaming ~8.5%, Health/Fitness ~8%, Tech ~7.5%, Beauty ~6.5%, Entertainment ~6%, Finance ~5.5%, Education ~4.5%. Sustained 4โ€“8% as reach expands is healthy.
  • Declining CTR is normal: the first 24 hours (warm subscribers) always score highest; CTR drops as the video reaches colder audiences. Don't panic-swap a thumbnail over that curve alone.
๐ŸšซWhat gets you suppresseddon't gamble the channel

Some thumbnail mistakes don't just cost clicks โ€” they cost distribution, or the channel.

  • Suppression triggers: nudity or suggestive content, violence / graphic imagery, vulgar text, and misleading packaging. YouTube now automatically compares what the thumbnail + title claim against what the video actually delivers โ€” bait that doesn't pay off gets suppressed, not just down-ranked.
  • The strike ladder: first violation = a warning that expires after 90 days; the same violation again within 90 days = a strike; 3 strikes = channel termination. Pornographic thumbnails skip all of that โ€” immediate termination, no appeal.
  • The safe play: shocking is fine, lying isn't. Use a real moment from the video, exaggerated honestly โ€” that also protects retention, which is what actually spreads the video.