Why can content not just be made in 16:9?

Because the surface is not 16:9. A mapped scenic build, a stacked LED array or a wall with holes in it has a canvas shape of its own, and authoring to the wrong shape means the picture is cropped, stretched or landing on structure. The canvas comes from the geometry, which is why content is blocked until previs delivers it.

How long should a loop be?

Long enough that the repetition does not register, short enough to render and store. In practice most beds run somewhere between eight and sixty seconds. What matters more than duration is whether the loop has an obvious landmark in it, a bright event that recurs is what the eye locks onto and counts.

What is the difference between playback content and generative content?

Playback content is rendered in advance and played as video; generative content is computed live and responds to input. Playback is predictable, cheap to run and impossible to change on the night. Generative reacts to tempo and to the operator, but consumes machine resource during the show and can fail live.

When does content need to be finished?

Far earlier than anyone plans for. Content is downstream of geometry and upstream of everything visible, so it absorbs every delay above it and passes none on. Productions that protect one date usually protect the content lock, because slipping it slips the show.

How is content organised for handover?

By layer role and energy, matching the way the VJ will use it rather than the way it was made. Naming that describes function survives showtime; naming that describes the project does not.

Content Creation for Live Shows

Abstract layered motion graphics in magenta and amber, arranged as separate depth planes. — Authored per layer · because that is how it will be played

Content creation for live production shares software with film and almost nothing else. The frame is not standard, the material must survive being seen a hundred times in one evening, it has to be legible through haze at fifty metres, and it usually has to be finished before anyone has seen the surface it will land on.

Authoring to a canvas that is not a rectangle

The first question is what shape to draw. A mapped set has a canvas defined by its unwrapped geometry; a stacked LED rig has one canvas per layer; a wall with scenic holes in it has a canvas with holes in it. Getting this wrong is not a stylistic error, it is a technical one, the content lands on structure, or falls off the edge, or arrives stretched.

This is why content creation is blocked behind previs. The canvas is an output of the model.

Loops that survive repetition

Almost all bed material loops, and the failure mode is always the same: an event in the loop that the eye can count. One bright flare, one distinctive shape crossing the frame, and by the third repetition the audience knows the length of the loop and stops seeing the content.

The defence is to avoid landmarks. Continuous motion without a signature event, several elements cycling at different periods so their combination takes far longer to repeat, and no hard cut at the loop point. Good bed content is close to featureless on its own and only becomes interesting layered.

The render pipeline

Rendering for a live rig has constraints that film does not impose. Output resolution follows the physical pixel map, which is frequently a strange number. Codecs must decode reliably in real time on the show machine, in parallel, alongside everything else. File sizes have to fit local storage on hardware that will be in a truck.

The practical result is that the render pipeline is validated on the actual show machine, early, with real files. Discovering a codec problem during load-in is a category of mistake that only happens once per career.

An irregular non-rectangular canvas floating in black: a shape assembled from several angled panels of different sizes, content flowing across all of them as one continuous field of light. — One field across many panels · not a frame, and not several

The canvas here is the hard part. Content authored to a 16:9 frame and then fitted to a shape like this reads as a picture that has been cropped by furniture; content authored to the shape reads as light that belongs to the object.

Which means the geometry has to arrive before the first frame is drawn. That single dependency explains most of what goes wrong in this discipline: not a shortage of skill or time, but a canvas that changed after the work against it had started.

Handing over to the operator

Content is not delivered as a folder. It is delivered as a populated clip bank, organised the way it will be used, by energy and role, not by project. Where content needs to react rather than repeat, it moves out of the render pipeline altogether and into real-time generation, and where it is played back it usually lands in Resolume.

Why it is the slowest thing in the room

Content is the only element that cannot be produced faster by adding people at the end. A render takes the time it takes, a loop that does not sit right has to be reworked rather than patched, and the material has to exist before anybody can rehearse with it.

It is also the element that absorbs every delay above it. Geometry late by a week does not cost a week here; it costs a week plus whatever re-authoring the change implies, and a scenic change after content has started is the single most expensive event in the schedule.

What it needs, and who receives it

It needs a frozen canvas from previs and mapping, a resolution and aspect from the LED specification, and a direction to aim at from art direction. It hands its output to the operator, who will play whatever the banks make easy, so the material has to carry the intent, not merely allow it.

The programs involved are under the toolchain, and terms used here are defined in the glossary.

Content creation for live shows: loops, canvas and the render pipeline