Studio · 2026-03-01 · 5 min
Ships Under Live Conditions
Software ships when it is merged. A show ships when the doors open, in front of people who paid, on hardware that was installed that afternoon, in a room nobody could fully test in advance. There is no rollback.
That single constraint shapes every decision we make, and it is the reason our process looks conservative to people who only build for screens.
The first rule is that unknowns have deadlines too. Screen resolution has to be confirmed at least seven days before the event. It sounds bureaucratic until you have rebuilt a composition at 2 a.m. because a LED wall turned out to be 2560 wide instead of 1920. Resolution, refresh rate and signal chain are not technical details to settle on site, they are inputs to the work.
The second rule is that setup time is part of the design. A club date needs one hour of setup and thirty minutes of video check. A festival needs two hours of setup and a full hour of technical rehearsal. If the schedule does not contain those blocks, the schedule is wrong, and no amount of preparation will fix it on the night.
Brightness is decided by context, not by preference. A LED wall indoors is one situation. A festival tent needs twenty thousand lumens as a floor. Outdoors, twenty to forty thousand. Getting this wrong does not produce a slightly dimmer show, it produces an invisible one.
The third rule is redundancy that a human can operate. On the 100 Violins build, alongside the TouchDesigner network there was a browser-based generator serving both as preview and as an NDI fallback. If the main machine had failed, the fallback did not require debugging a node graph under stage lights. That is the test we apply: can this be recovered by a tired person in a dark room in under two minutes.
The fourth rule is that the runbook is a deliverable. Not documentation written afterwards, a document opened on the day, listing the files, the network, the order of operations and what to do when something does not respond. Djerba ran on execution drawings, a device-by-device interaction map and a complete technical inventory down to the sensor model and the resistor value. That level of specification is not excessive for a permanent installation with sensors, DMX and nine microphone inputs. It is the minimum.
None of this is about lowering ambition. It is about making sure the ambitious part is the thing the audience sees, and not the thing the crew is quietly fighting backstage.