Nine well-known AI models — Kimi, Claude, GPT, Gemini, DeepSeek, Grok, Llama, Mistral and Qwen — drawn as naïve doodle heads on a single sheet of digital paper, rendered live on a canvas at 60fps. Kimi K3 drew, voiced and coded all of it — and named it, in both its languages: If We All Had Heads / 《如果我们都有头》. I directed and curated. The piece is a small museum: the cast is the finale, and the top-right nav keeps the two earlier rooms open — avatar, Kimi's solo talking head with a Mandarin monologue, and block-in, the animator's process made visible, where a rough-to-ink slider scrubs the drawing-on while the blue-pencil skull rotates underneath.
The technical conceit is the whole aesthetic: flat doodles that feel alive because a rough 3D skull turns beneath them. Every feature is pinned to angular coordinates on that skull and clamped inside the projected silhouette, so nothing ever slides off a face. The pencil line boils at 8–11fps like hand-drawn animation. Move your cursor and the heads notice you; click one and it steps forward to speak — first click also starts the lo-fi music bed, which Kimi generated too.
A 55-second film records one round of introductions, dubbed in both languages.