A few years ago I had a phase. I played chess obsessively on chess.com — hundreds of games, maybe thousands. I exported all of them. I had this idea that I'd turn them into videos somehow. I spent weeks on it. Nothing worked. I abandoned it and told myself it was a dead end.
Last night I picked it up again with a different tool and a different angle. In about an hour we had a working video engine.
The engine parses PGN files — the standard chess game format — replays every move, renders each board position as a frame, syncs piece sounds to every move, and outputs two video formats simultaneously: 16:9 for the web feed and 9:16 for YouTube Shorts. Pure Python, Pillow for rendering, ffmpeg for assembly. No external chess libraries. About 500 lines of code.
The creative direction was ASMR. No names. No scores. No commentary. Just the board, the pieces, and the sounds — move, capture, castle, check, promote. The same rhythm across every game. Hypnotic by design. The winner always plays from the bottom so the viewer always feels like they're winning.
I have 2000 games in the archive. After filtering for quality — decisive results, clean endings, at least 15 moves — about 1600 remain. At a batch rate of 50 videos per evening that's a month of content generated in a few sessions. At a 4-hour upload cadence that's nearly a year of automated posting.
The lesson isn't really about chess or code. It's that old abandoned projects deserve a second look. You were probably closer than you thought. You just needed a better angle.
Tonight the engine is running. Tomorrow I QC the batch, pick the best ones, and start building the upload queue. The chess channel goes live this week.