yunyiliu/cvpr-2026-poster-skill

This project is a skill you install into an AI coding agent like Claude Code or Codex, and it helps you turn a CVPR 2026 conference paper into a print ready poster. Instead of using PowerPoint or fighting with templates, you talk to the agent in plain language and it builds the poster for you. Once installed, you point the agent at your paper's Overleaf folder. It reads the LaTeX source to pull out the title, authors, affiliations, abstract, and figures, then uses that information to fill in an editable poster made of HTML. It also tries to automatically find and download your institution's logo from your school's website, and it adds the official CVPR 2026 conference logo. After the first draft is built, you keep refining it just by describing changes in chat, things like moving a section to a different column, shrinking a logo, or trimming a card's text down to a shorter summary. You can also edit the poster directly in your browser: dragging dividers to resize columns and rows, dragging cards to swap their positions, and scaling all the fonts up or down at once. When you are happy with the layout, you save it, which downloads a configuration file with your changes. The final step converts that layout into a PDF sized exactly for either the main poster track at 84 by 42 inches, or the smaller workshop track at 42 by 21 inches, using a headless version of Chrome to do the rendering. That PDF can be sent straight to a print shop. To use this, you need Google Chrome installed for the export step, and Python 3, which comes preinstalled on macOS and most Linux systems. Everything else the tool needs, like the LaTeX reader and logo fetcher, ships inside the project itself. There is also a manual path for people who would rather run the individual Python scripts themselves instead of going through an agent conversation, useful if you want more direct control over each step. The project is released under the MIT license, so it can be freely reused, including for commercial purposes.