Create a conference poster for NeurIPS, ICML, ICLR, or CVPR by filling in an HTML template and exporting a print-ready PDF.
Automatically check your poster layout for column overflows or sizing issues before submitting to a venue.
Use Posterly as an AI coding agent skill inside Claude Code to walk through the entire poster creation workflow.
Requires Python 3.10+, Playwright, and system PDF utilities, MathJax loads from CDN by default but can be used offline with a local copy.
Posterly is a tool for creating academic conference posters. Instead of using LaTeX, which is the traditional approach in research, it builds posters as a single HTML and CSS file and then uses a headless web browser to convert that file into a print-ready PDF at exact dimensions for specific venues like ICML, NeurIPS, ICLR, and CVPR. The reasoning behind choosing HTML over LaTeX is mainly about the editing experience. Changing layout or styling in HTML and refreshing a browser takes seconds. Doing the same in LaTeX means recompiling, reading error logs, and reopening the PDF. Web layout tools like Flexbox and CSS Grid also offer more flexible positioning than LaTeX poster packages, and since the poster is a structured HTML document, automated checks can read the geometry programmatically to detect column overflow or other layout problems instead of relying on visual inspection. The project ships as a coding-agent skill, meaning it is designed to be invoked from within an AI coding tool like Claude Code or Codex. When you run the slash command inside your agent, it walks through the steps of picking a venue, selecting a template, filling in content, running sanity checks, and rendering the final PDF. It can also be used directly from the command line without an agent. Four command-line tools handle the quality checks: a preflight that validates the HTML structure, a measure tool that checks column and canvas sizing, a polish check for content quality, and a final verify step that inspects the rendered PDF dimensions against the venue spec. Three poster templates are included (four-column landscape, hero landscape, and two-column portrait), along with three real published conference posters as worked examples. Installation requires Python 3.10 or later, the Playwright library for browser control, and the system utilities for reading PDFs. Math in posters is handled by MathJax, which loads from a CDN by default but can be swapped to a local copy for offline use. The license is MIT.
← chenruishuo on gitmyhub — every repo by this author, as a profile.
Verify against the repo before relying on details.