How It Works · Hybrid Neural + Vector Architecture
End-to-end image diffusion models garble speech-bubble typography. xkcd-artist masks dialogue out of 226 XKCD panels during NVIDIA L4 training, executes a 1.72 MB Int8 quantized 5-head ONNX planner directly in your browser via WebAssembly, and composites crisp xkcd-script vector paths, speaker leader lines, and hover title text.
01 · Speech-Bubble-Masked Training
Dialogue in the top 36% of 226 XKCD panels is masked out before NVIDIA L4 fine-tuning alongside 891 ExplainXKCD scripts, isolating pure stick-figure and stage-prop ink geometry.
02 · In-Browser ONNX WASM
03 · Vector Compositor
4x hand-wobbled #0A0A0A pen paths, head-crown leader lines, CC-licensed xkcd-script lettering, and Randall-style hover title text render at infinite SVG resolution.