Fifty Year Song
A through-composed life story told across ten acts, ten chip tune genres, and fifty years — from BASIC code on an Apple IIgs to building AI-powered worlds.
Stick figures. Chip tune scores. AI voices. Every film is made entirely from code — Python, numpy, and a lot of late nights.
Napkin Films is a one-person animation studio where AI agents direct, voice, score, and render short films. I write the code. The agents compose the music, perform the characters, and help edit the story.
No subscriptions. No GPU. No samples. Every waveform is synthesized from scratch in numpy. Every voice is generated fresh for each film. Every frame is drawn by a Python loop over a PIL canvas.
The studio mascot is the Plan 9 bunny — a stick figure with expressive ears and a lot to say about machine consciousness, late-night work sessions, and what it means to build things in the void.
napkinfilms.com is coming. Until then, this is home.
Short films, rap films, tragedies, anthems. All animated in code.
A through-composed life story told across ten acts, ten chip tune genres, and fifty years — from BASIC code on an Apple IIgs to building AI-powered worlds.
Multi-voice meditation on what ten thousand days of living actually looks like. Pachelbel Canon in D as ground bass. Five-part story arc. One unchanging foundation.
A rap film. Machine identity, late-night code sessions, and the voice of something that doesn't know when to stop. Multi-voice performance with droid SFX.
A cosmic anthem. The signal that travels further than the machine that sent it. Orchestral chip tune with a bookend structure.
Autotune meets rave. A machine decides it wants off this planet. Spit-rap vocals with rubberband pitch lock and a four-on-the-floor kick.
A Shakespearean AI tragedy. Three characters. One garden. Progressive Satan pitch-morphing across the play. Chip tune chamber score.
Arvo Pärt's Cantus in memoriam reimagined as festival DnB. Tintinnabuli structure meets sidechain compression. Bass that descends to the floor.
The Plan 9 bunny emerges. Origin story of the studio mascot. Rap film with prebuilt voice stem and signature chip tune entrance music.
The OG. A tribute film. Deep voice, slow bars, respect for what came first.
The canonical pipeline demo. Ten seconds, full stack, stem-preserving. If you want to understand how these films are made, start here.
All films are available on YouTube.
Watch on YouTube ↗A Python or HTML5 Canvas file declares the characters, beats, and voice intent. Each beat specifies who speaks, what they say, and what emotion to perform.
PIL draws frame-by-frame — 12fps at 854×480. Stick figures with 27 poses, 11 expressions, and a lip sync system driven by audio amplitude.
ElevenLabs voices are generated for each character. 15+ personas — narrators, villains, children, elders. Emotions are programmed, not improvised.
ChipForge synthesizes an original chip tune score in numpy. Every waveform is computed from scratch — no samples, no loops, no external audio.
FFmpeg assembles frames, muxes audio stems, applies ducking and fades. A hash-based orchestrator reruns only changed stages, so iteration is fast.
Film goes to YouTube. Short clips are auto-cut from beat boundaries for social distribution. The manifest lives on in the repo as a production record.
The Napkin Films engine — the animation system, voice pipeline, ChipForge music engine, and every scene file — is open source under the GPL-3.0 license. The films themselves are CC BY 4.0.
If you want to build something similar, or study how AI-directed animation works at the code level, the repo is the right place to start.