denniswei9898/social-cards-engine — explained in plain English
Analysis updated 2026-05-18
Turn an article or topic into a ready to post Instagram carousel of image cards.
Swap a single brand pack file to give the same engine a completely different visual style.
Have an AI reviewer score and reject cards that fail structure or readability checks.
Automatically find and attribute copyright safe images or memes for your cards.
| denniswei9898/social-cards-engine | av2xn/doan | chandar-lab/semantic-wm | |
|---|---|---|---|
| Stars | 30 | 30 | 30 |
| Language | Python | Python | Python |
| Setup difficulty | moderate | hard | hard |
| Complexity | 3/5 | 5/5 | 5/5 |
| Audience | vibe coder | developer | researcher |
Figures from each repo's GitHub metadata at analysis time.
Requires macOS with Chrome installed for the headless screenshot rendering step.
social-cards-engine is a Claude Code skill for creating Instagram style carousel image cards. The idea is that one engine can produce many completely different visual styles just by swapping out a brand pack, a single file describing colors, fonts, sizing, and voice for a given brand, so you can go from a navy poster look to a casual meme style to a premium food brand look without redoing any underlying setup. The project splits its parts into three roles: the engine itself, which turns HTML into a screenshot and then a row of properly sized PNG cards, the brand pack, which is the swappable recipe describing how a particular brand should look, and a built in AI reviewer called a joker that scores every finished card and rejects ones that fail, so a person is not the only one checking quality before publishing. The workflow starts by picking a brand pack, optionally mocking up a single test cover card first to confirm the look before mass producing the rest, then splitting your topic or article into cards, rendering them through a headless Chrome browser, and sending the result to the reviewer for a pass or fail check on things like structure and humor. If it fails, cards get revised and resent for review, up to three rounds before a human is asked to step in. The project also includes starter templates for different visual styles, a system for finding and attributing copyright safe images or memes from public domain and creative commons sources, and a script that checks brand logos and mascots for stray white edges before they get used on cards. It works alongside two optional companion skills, one that checks layout quality such as centered images and clean spacing, and another that applies research based rules about what makes content get shared, plus semi automatic posting to platforms like Instagram and Threads. The project is aimed at solo content creators, brand editors, and agencies handling multiple clients, and it runs on macOS with Chrome, written in Python, and released under the MIT license.
A Claude Code skill that turns one article or topic into a branded Instagram carousel of image cards, with an AI reviewer checking quality before publishing.
Mainly Python. The stack also includes Python, HTML, Chrome headless.
Setup difficulty is rated moderate, with roughly 30min to a first successful run.
Mainly vibe coder.
This repo across BitVibe Labs
Don't trust strangers blindly. Verify against the repo.