Bar-Ma started as a question about timing. Most screen content we make for live events is rendered in advance and played back against a cue list. That works, and most of the time it is the right answer. It also means the picture cannot respond to anything happening in the room. Bar-Ma is the other approach: a system that draws frames while the show runs. It works on two things, the bar as a unit of rhythm and the interval between bars. The bar sets the structure. The interval is where the picture changes.

What did not hold up

Three things broke. The audio-reactive layer was the first. In the studio it read a clean feed and behaved well. In a room with a PA, a crowd and an open mic it read noise instead. The picture twitched on coughs and applause. We dropped it in favour of a tempo the operator sets by hand, with timecode as the fallback. Resolution was the second. Detail that looked precise on a desk monitor turned to mush on a wide LED wall with a coarse pixel pitch. The fine line work we liked most was the first thing to go. Control was the third. We had exposed more parameters than anyone can steer during a keynote. Interesting to sit with, useless to run.

What survived is smaller than what we started with. Bar-Ma now runs from a short preset rack the operator can move between. Every look is built against the real canvas ratio before anything else is decided. We also render a still sequence of each preset, so there is something to cut to if the generator stalls. It is not a product and it has not gone into a client show. It sits in the lab. We take pieces out of it when a screen brief needs content that moves with the room rather than ahead of it.