ExcaliAI - New AI Capabilities Have Arrived to Obsidian Excalidraw

Excalidraw’s new ExcalAI script turns hand-drawn notes and wireframes into AI-generated outputs inside Obsidian—mind maps, diagrams, and even code—while also offering automated feedback on the content of what’s drawn. The core payoff is speed: paste text or upload a drawing, run ExcalAI, and get a structured visual artifact (or code) that can be refined with another prompt or a rerun when results glitch.

The workflow starts with ExcalAI being installed as a script in Excalidraw (via the Obsidian Tools panel → “Install Script” → “ExcalAI”). Using it requires a separate OpenAI API key and some account balance. The presenter notes that typical generations cost around three cents, but the vision features require at least $5 in the OpenAI account. Even with correct setup, outputs can occasionally fail or return unusable material, so rerunning the generation—sometimes with a tweaked prompt—is part of the practical routine.

One demonstration uses a page about “the art of close reading.” After switching Excalidraw’s output to Mermaid and running ExcalAI, the system produces mind maps that correctly identify the topic and then generates follow-up “feedback” questions. The feedback focuses on precision and accuracy—whether precision always enhances understanding—and on how (or whether) depth of an idea can be determined when breadth and depth are both in play. The key takeaway is that ExcalAI can generate reflective prompts tied to the semantics of the text or drawing, not just a visual summary.

A second feature converts a blog post into a mind map. The user copies Thiago Forte’s “second brain” blog post text, chooses Excalidraw’s “text to diagram,” pastes the content, and generates a structured diagram. The mind map summarizes organizational and training concepts—setting goals, deploying training, and improving productivity, communication, learning, and project management—then presents them as a visual hierarchy. The same text-to-diagram capability can also produce other structures such as process flows and argument maps.

The most hands-on segment shows wireframe-to-code. A calculator wireframe is converted into HTML using ExcalAI’s “wireframe to code” output type. The system sometimes generates broken code, but retrying can yield working results. A quick test illustrates this: “12 * 3” fails in one generated version, while “20 + 1” works; a later generation produces a calculator where “12 * 3” returns “36.” Excalidraw also includes a built-in “wireframe to code” path (with a slightly different prompt), and both routes can output source code that can be copied to the clipboard.

Overall, ExcalAI is positioned as a toolkit for building AI-enabled Excalidraw applications: the script can be installed, inspected, and used as a reference for creating new functions, with release notes listing added capabilities. The practical message is clear—expect iteration, but the payoff is turning notes and sketches into structured thinking and usable artifacts inside your knowledge workflow.