thumbnail-maker
Skill1 starsChanged 6 days agoMITView on GitHub
Make YouTube and social thumbnails (1280x720, plus 1080x1920 Shorts/Reels and 1200x630 link previews) with Pillow - a large auto-fitted title with one highlighted keyword on a clean flat or blurred-photo background, plus either a real screenshot with a hand-drawn red circle and arrow (tech and tutorial style), or a person photo/cut-out that can stand in front of the text. Checks readability at YouTube search size, the squint test, and YouTube's timestamp corner, and writes a preview sheet. Use when the user asks for a thumbnail, cover image, video title card or gig image.
- Installs packages
3 other projects use a version of this.
Adopted by
Most recently taken up by
- Mikefluff/skillstheir own version, first seen Aug 8, 2026
- Retired-Projects/ai-marketing-kit-publictheir own version, first seen Jul 30, 2026
- crealwork/ai-marketing-kittheir own version, first seen Jul 17, 2026
Other versions3 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 3
- Mikefluff/skills★ 26same name
- crealwork/ai-marketing-kit★ 18same name
- Retired-Projects/ai-marketing-kit-public★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository25 other files it gives its agents.
- bash-checkpointSkill
- brave-cdpSkill
- cf-backupSkill
- cf-pages-deploySkill
- claude-md-doctorSkill
- claude-safety-auditSkill
- claude-sandboxSkill
- claude-token-auditSkill
- cws-extension-checkSkill
- env-doctorSkill
- event-recap-videoSkill
- freelance-proposalSkill
- gif-makerSkill
- itch-html5-packageSkill
- link-preview-checkSkill
- mcp-auditSkill
- package-guardSkill
- pwa-checkSkill
- read-guardSkill
- readme-upgradeSkill
- release-notesSkill
- secret-commit-guardSkill
- session-handoffSkill
- site-qaSkill
- visual-diffSkill
The file
We split this skill into the rules it gives an agent
3 rules. About 0 are common, 3 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 37
Keep the bottom-right corner clear: YouTube draws the video length there.
- Distinctive: line 48
No background removal: pass a PNG that is already cut out (remove.bg, Photoshop, macOS "Copy Subject").
- Distinctive: line 54
No face detection: crop photos so the face isn't near an edge.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
thumbnail-maker
scripts/thumbnail.py (with scripts/fonts.py next to it). Python 3.9+ and pip install pillow.
# tech / tutorial: a real screenshot, the important line circled, an arrow from the title
python thumbnail.py --title "Claude almost deleted *everything*" --style dark --shot terminal.png --circle 0.19,0.17,0.18 --arrow --out t.jpg
# talking-head: a cut-out person in front of the title
python thumbnail.py --title "I tested *every* AI coding tool" --style dark --photo me-cutout.png --out t.jpg
# photo background, Shorts size
python thumbnail.py --title "A day as a *dev*" --bg desk.jpg --style dark --size 1080x1920 --out short.jpg| Option | What it does |
--title | 3-5 words reads best. *word* marks the keyword (one box or color). Uppercased unless --keep-case. |
--style | dark (near-black + yellow), light (off-white + red), or the louder pop, red, green, tech. |
--look | clean (default): flat color with a soft vignette, or --bg photo blurred and darkened for contrast. burst: the older gradient + sunburst look. |
--shot img | A real screenshot as the main visual. Wide crops (one or two lines of a terminal or UI) go full-width under the title; taller ones sit beside it. Scaled up to fill the space. |
--circle x,y,r | Hand-drawn red oval on the screenshot. x,y = center and r = half-width of the thing, all as fractions of the screenshot. |
--arrow | Hand-drawn arrow from the title to the circle (routes around the title when there's no room above). --mark-color changes the red. |
--photo img | Person photo. A PNG with transparency (a cut-out) is outlined and drawn in front of the title edge for depth; a normal photo becomes a bordered card (--photo-shape circle, --tilt). |
--subtitle, --emoji, --accent, --highlight color, --size, --align, --font | As before: a small second line, an emoji sticker, keyword color, colored keyword without a box, output size, alignment, title font. |
Design rules (from what top thumbnails do in 2026)
- One idea, 3-5 words. The title complements the image; don't describe it.
- Show the real thing. A cropped screenshot of the actual error, result or UI beats generic art. Crop tight to the line that matters and circle it.
- Faces: a cut-out with a clear expression, large (40-60 % of the height). Don't put text over the eyes.
- Two colors. Flat background plus one accent. No sunbursts, emoji walls or gradient text; they read as template or AI-made.
- Keep the bottom-right corner clear: YouTube draws the video length there.
Checks (printed, and on <name>.preview.png)
- Title letter height at 168 px wide (YouTube search) under 8 px → warning.
- Effective contrast of the title (fill vs background, and its outline) under 3:1 → warning.
- Squint test: the title area blurred to 48 px wide must still stand out from its surroundings.
- Timestamp corner: title reaching the bottom-right 20 % x 16 % → warning.
- Keyword box blending into the background, keyword vs box/outline contrast, more than 6 words. Exit code 1 on any warning. LOOK at the preview sheet before using the image.
Limits
- No background removal: pass a PNG that is already cut out (remove.bg, Photoshop, macOS "Copy Subject").
- Circles and arrows are placed from your
--circlenumbers: check the preview and adjust x/y/r if the oval misses. - Fonts: Impact for titles (Windows/macOS; on Linux install
ttf-mscorefontsor setFONT_TITLE=/path/Anton-Regular.ttf). Override withFONT_TITLE,FONT_BOLD,FONT_EMOJI. - Devanagari (Hindi/Nepali) titles and subtitles switch to Mangal Bold / Nirmala UI automatically (
FONT_DEVANAGARIoverrides). Pillow without raqm (stock Windows) can't join conjuncts or place the short-i matra, so that text is shaped by ffmpeg's drawtext (HarfBuzz) word by word and composited in; mixed English + Nepali titles work. With no ffmpeg built with libharfbuzz on PATH the check warns instead, so read the preview. - Characters the font can't draw (an emoji typed into
--title, for example) are reported as a warning instead of silently becoming boxes. - Emoji: single-codepoint only (Pillow without libraqm can't join ZWJ sequences).
- No face detection: crop photos so the face isn't near an edge.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.