Playing the Picture Live

Every piece this week kept circling the same idea: what if you didn't plan the picture, what if you played it? Some of these people did it on hardware in the '70s with no undo. Some are doing it now with a generator and a deadline. Same nerve either way. You commit to the take, the mistake shows up, and sometimes the mistake is the hook.

Tom DeFanti and Dan Sandin on Playing Images

Part 1 of Peter Bauman's interview with Tom DeFanti and Dan Sandin for Le Random. It covers how the Electronic Visualization Lab at UIC started: DeFanti's real-time GRASS graphics language, Sandin's Image Processor (an analog video synth he thought of as a visual Moog), and how the two got wired together into something you could play live, like a band.

This one made me want to call Flint at 2am. Sandin looks at RGB video and basically hears audio, just at a way higher bandwidth. It's all signal, and the instrument is really a frame of mind. Then they took that instrument into inflatable structures and threw Electronic Visualization Events. They built their own rooms so they could make their own rules. That's DIY festival energy fifty years early.

The part I love most is that the "errors" are the style. Spiral PTL runs on temporal aliasing. Circle 9 Sunrise comes from abusing the way a Vector General drew circles. That's glitch at the hardware level, before anybody had a word for it. And then they gave the whole thing away. No patents, just documentation, and Phil Morton's "copy-it-right" ethic. They shared the instrument itself, not only what came out of it.

Real time is a great teacher. I mean, you do something and you make a mistake and you know about it, right?

Calculated Movement | Larry Cuba

Georg Bak interviews computer-animation pioneer Larry Cuba for Right Click Save. Cuba made First Fig at NASA's Jet Propulsion Laboratory in 1974, worked with John Whitney Sr on Arabesque, and did the Death Star briefing graphics in Star Wars. He has made four films in fifty years, and these days he's building his own programming language, OSKAR.

Four films in fifty years. No storyboard, ever. Each film comes out of the experiments. He starts with a dot or a little cube, repeats it, transforms it, and listens to what it wants to be. In the '70s his feedback loop was weekly: write the program, run it over the weekend at JPL, pick up the film Monday, watch it at Whitney's house Tuesday. One look at your take per week, and he still finished. Picture that the next time you hit undo forty times on one snare.

What really got me is that he won't let the music analogy off easy. Everybody in generative art waves "visual music" around like a flag. Cuba actually pushes back on it, and as a music guy I felt seen and corrected at the same time. That's a good feeling, honestly.

Music works because it follows certain rules. Visual music is not music. Visuals have rules of their own and the trick is to find out what they are.

Adrian Brück Made a Robot the Narrator of His AI Mockumentary

Adrian Brück, who posts as odngr on Higgsfield, made "And Where Are We Going?" It's a mockumentary under six minutes, shortlisted for Higgsfield's Global Film Festival. A robot narrator flatly watches its Neanderthal buddy, who is obsessed with building a boat. Brück built it with Seedance 2.5, GPT Image 2, ElevenLabs, and Suno over 522 generations, and left the whole messy workbench public.

This is the smartest constraint move I've seen in AI film. Synthetic voices sound flat, and in a human character that kills the scene. So he gave nearly all the lines to a robot, where flat is the character. He didn't fix the weakness, he cast it. It's like writing a melody right around a singer's break because the break is where the feeling lives. The nature-doc framing helps too, because a long lens is supposed to feel distant, so the uncanny stuff reads as a choice instead of a flaw. Then the Neanderthal gets one word, "Boat," and knocks the boom mic out of frame. That's real comic timing.

The hat was my way of showing how we humanize technology to make it graspable.

Brent Bonacorso Answers Five Questions About His New Short Film "Query"

Director Brent Bonacorso talks to Stash about Query. The premise is that AI chatbots are secretly staffed by thousands of people typing back answers. He dropped his usual script and storyboards and built it in order, one shot at a time. He calls the process "vibe directing."

This is a guy with decades of visual discipline who decided to find the film the way you find a song in a session. He keeps the look tight by baking lenses, light, and texture into the prompts, and lets the story stay loose. Rigid on sound, loose on structure. And then there's the duck, which I'm going to be thinking about all week:

For instance, the giant duck started as a background joke. Then somehow it followed the protagonist through the movie and eventually became crucial to the resolution of the story. I had absolutely no idea that was going to happen when I started.

That's the detour becoming the song. You only get that if you leave room for it.

---

What's stuck in my head is Sandin turning a knob in 1972 and Bonacorso chasing a duck in 2026, both finding out what the thing was by playing it. I want more work that's brave enough to be a take instead of a plan.

Mac