Glitch visuals in Max

Accina Studio — Wed, 14 Oct 2026 00:00:00 +0000

Some of the visuals behind the sets come out of patches like this one, built in Max from Vizzie modules and a little Jitter.

Roughly, the signal goes like this:

Sound in. An audio file plays inside the patch. AUDIOSPLITTR splits it into frequency bands, and AVERAGR smooths each band into a steady control signal.
Movement. 4OSCIL8R adds slow oscillators, and RANGR scales each signal to the range the next module expects.
Picture. CROPPR and INTERPOL8R crop and stretch the video. In Jitter it becomes a texture on a 3D grid, pushed up and down by the sound.
Out. The finished picture can be sent on to other video software through Spout.

Patching on the laptop, with the output on the big screen. Every knob change shows up instantly, which is half the fun.

New to Vizzie? The Vizzie tutorial starts from the beginning.

QR codes you can hang on a wall

Accina Studio — Wed, 14 Oct 2026 00:00:00 +0000

A QR code is usually the ugliest thing on a flyer. I wanted one that is the artwork instead: an image you look at for the colours and shapes, which still opens a link when you point a phone at it.

Try it: point your phone at the screen. A dedicated QR-scanning app will most likely read all of them. They have all been tested with the built-in camera app on a Google Pixel 9, but some phones’ camera apps are pickier with artistic codes like these, so if yours does not react, try a scanner app.

How I make them

I build the images in ComfyUI, a free, open-source program for running image-generation models on your own computer. Instead of a single text box, ComfyUI is a modular environment: you connect small building blocks (nodes) into a graph, one for the model, one for the prompt, one for the sampler, one for the output, and so on. That makes every step visible and adjustable, and a finished graph can be saved and reused as a recipe.

The key block is a ControlNet. A ControlNet lets you steer an image model with a second picture besides the text prompt, for example a pose, a depth map or, here, a plain black-and-white QR code. While the picture is being generated, the ControlNet keeps nudging it towards the light and dark pattern of the code, so the artwork grows around the code instead of covering it up.

The hard part is the balance. Too little pull from the ControlNet and the picture looks great but no phone can read it. Too much and it is just a QR code with a filter on it. For every prompt I render a ladder of versions from loose to strict and keep the most artistic one that still scans.

A design is never finished until a real phone has scanned it, in normal light, from a normal distance. Everything runs locally on my own machine: no online service sees the links or the images.

Get your own

I make custom codes to order: for a release, an event, a flyer, a shop window or a business card. Tell me where the code should lead and what it should look like, and you get a set of scannable artworks to choose from. Contact me to buy one.

Where this is going

Later I want to offer a generator you can use yourself: type a link, describe the image, get a set of scannable artworks back. That needs a graphics card in the cloud, which this website’s hosting cannot provide, so it is a project for later.

Accina Studio - Lab

Glitch visuals in Max

QR codes you can hang on a wall

How I make them

Get your own

Where this is going