cy-carousel

cy-carousel is a carousel Skill open-sourced by the GitHub account chengyi-ai, and the README marks it as working with both Claude Code and Codex. Give it a concept, a paper or a product, and it produces a cover plus 10 inner pages at 1440×1920, delivered with a title, caption, pinned comment and source list. The process starts with choosing a topic: it lists a few candidates and compares their material and factual risk, then traces every number and quote back to the original paper or primary source and records them in a fact sheet. It uses only public-domain and openly licensed material, logging each item's link, license and the page it appears on. The layout is one fixed look: a dark or light paper background, two-color brush-script headlines, enlarged excerpts from papers circled in red, and old photos or real screenshots as the main visual. The README says the format was tuned by testing it on one AI topic after another, such as ELIZA, AI hallucinations and Clever Hans. What sets it apart is that fact-checking and image licensing are part of what gets delivered, with one command checking its rules and text overflow; the scripts only write files and never publish anything.

Who it's for:creators who make AI explainer carousels for social media, especially those willing to get the facts and image sources right.

Before you start: it needs Python 3.10 or later plus dependencies such as Pillow, numpy and PyMuPDF; install it in Claude Code or Codex and start a new conversation. Finding words in scanned book pages works only on macOS, and the rest also runs on Windows. The layout is designed for Chinese text and has a fixed style. The AI helps with fact-checking and finding images, but you still need to verify the facts and each image's license before posting. Code and docs are MIT, and the four bundled fonts are under the SIL OFL 1.1.

Based on the project's public documentation. Not tested by AI Inspiration World.