@di-atomic/thumbnail

September 24, 2026

Name: @di-atomic/thumbnail
Tagline: A video and its title in, the click image out: a YouTube A/B set of real ideas, covers for Shorts and Reels, and the poster frame, each graded where it is shown smallest.
Hero subhead: A thumbnail that looks great in the editor is judged 200 pixels wide in YouTube's sidebar, with a time stamp over its corner. This skill plans two or three variants that each lead with one idea, grades every image at that size, and fails a set whose variants look alike. Every check prints a number.
Who it's for: Founders and marketers who publish on YouTube, and who have run a thumbnail test that came back with no winner.
Version: 0.1.0
Tier: beta
Path type: guidance-only
Install command: opvs-skills install @di-atomic/thumbnail --to ./skills-cache
Trigger phrase: make 3 thumbnails to A/B test for this video
Marketplace URL: https://market.opvs.ai/v/di-atomic/thumbnail
Blog URL: https://di-atomic.com/blog/shipping-di-atomic-thumbnail
Stat 1 value: 200 px
Stat 1 label: the width YouTube's sidebar shows your thumbnail in a laptop window
Stat 2 value: 26% x 21%
Stat 2 label: of that thumbnail sits under the duration badge. Words there are lost.
Stat 3 value: 0 to 1
Stat 3 label: fingerprint distance between a thumbnail and its recolour. Real pairs score 20 or more.
Stat 4 value: 1 in 10
Stat 4 label: of 941 niche thumbnails had nothing readable at 200 px
What headline: What you actually get
Cap 1 icon: 🧪
Cap 1 title: A YouTube A/B set
Cap 1 desc: Two or three variants for Test and compare, each leading with one idea: promise, subject, proof object, emotion, or face against object. Each one carries a one-line hypothesis, so the result teaches you something.
Cap 2 icon: 📱
Cap 2 title: Covers for Shorts, Reels, TikTok
Cap 2 desc: One cover with the words kept where the platform leaves them visible. YouTube lays its Shorts title over 74% to 88% of the cover's height, so your words stay above it.
Cap 3 icon: 🎞️
Cap 3 title: The right frame 0
Cap 3 desc: A settled, non-black frame, picked from a contact sheet and handed to @di-atomic/video-editor, which bakes it in and reads frame 0 back to prove it.
Showcase headline: What the gates see that your editor hides
Showcase before caption: Judged small: in a laptop window YouTube's sidebar shows your thumbnail 200 pixels wide, and the duration badge covers its bottom-right corner.
Showcase after caption: Ideas, not shades: a recolour looks identical at thumbnail size, so the set only passes when every pair of variants looks different.
Worked title: A real set, a recolour, and a poster
Worked example: 3 variants on test backgrounds, typeset headlines verify-thumbnail A PASS 19.0' @200px 21.00:1 badge clear verify-thumbnail B PASS 23.0' @200px 21.00:1 badge clear verify-thumbnail C PASS 18.8' @200px 21.00:1 badge clear verify-variant-set PASS A-B 27 A-C 31 B-C 40 (every pair >= 16) A and a recolour of A verify-variant-set FAIL V3 A-A2 0 · V4 A2 contrast 3.67:1 4 s clip, fades in from black, cut at 2 s pick-poster-frame PASS frame 0 today black 100% · 3 of 40 frames qualify
Worked framing: Output re-run from the published bundle on test backgrounds generated with ffmpeg. No face and no AI render: that path is the one untested part of this release.
Rel 1 icon: ✂️
Rel 1 name: @di-atomic/video-editor
Rel 1 desc: Bakes the poster as frame 0 and reads it back to prove it.
Rel 1 url: https://di-atomic.com/skills-video-editor
Rel 2 icon: 📊
Rel 2 name: @di-atomic/cro-optimizer
Rel 2 desc: Tells you whether a test win is real before you believe it.
Rel 2 url: https://di-atomic.com/skills-cro-optimizer
Rel 3 icon: 🎬
Rel 3 name: @di-atomic/video-director
Rel 3 desc: Turns a script into the shot list for the video your thumbnail sells.
Rel 3 url: https://di-atomic.com/skills-video-director
CTA headline: I wanted every video to get a fair test, not three shades of one guess.
CTA description: thumbnail is one skill in the system behind Di-Atomic, the marketing and compliance agency that runs cognitoAI, SpiderIQ and OPVS. Guidance-only, so there is no backend to provision. Your agent runs the scripts in its own shell with node and ffmpeg.
Upstream license: No third-party code. Ideas credited in PROVENANCE.md: the frame-0 poster idea from /brag (MIT, reference only) and dHash as a described method.