goopCodecs · databending codecs for Sgueltch

position Sgueltch / goopCodecs

Sgueltch Poetics

Sgueltch is the larger practice of making digital damage behave less like a grid and more like matter: silt, lesions, slime, cells, threads, fog, spread.

goopCodecs is the file-format branch of that practice. Instead of applying an effect to a normal image, it genuinely translates the image first into a new codec, which allows for more visually meaningful databending. Databending ordinary file formats can only go so far in terms of what they can result in, because they are limited by the shapes and logic of their codecs. If we want new visual effects from databending, we also need new ways of encoding data.

test what makes one good?

Databending Criteria

A good codec made for databending should allow for more meaningful moves in databending. It should keep a small protected header, leave the body vulnerable, and make many edits produce visible consequences.

Surfacehex view and mojibake/text view are both valid editing surfaces.
Tolerancetruncation, stale counts, and body misalignment should damage the specimen rather than immediately fail.
Morphologydifferent byte classes should have different visual identities: position, colour, scale, delta, frame length.
Cascadesame-length edits should be contained; length-changing edits should create downstream reinterpretation.

collection current working codecs

What the collection does

Mainstream formats usually optimize for storage, speed, or invisible loss. goopCodecs uses the codec as a practical instrument: the encode/decode process determines what kinds of damage can happen and what those errors look like.

The main operation is simple: import or generate a source image, encode it into the codec's native body, apply non-destructive damage or edit the bytes directly, then decode the altered file back into an image.

ooid

.ooid

Gaussian blobs → round loss

A still, colour-field, and video codec based on layered anisotropic Gaussian blobs. Images are stored as records for position, colour, size, orientation, and alpha, then composited in record order.

Corruption produces swelling blobs, relayering, fog, teleports, recolouring, and soft lesions. Video uses independent ooid frames in an OOIV container, so damage can stay frame-local or be pushed across boundaries.

Open the blob codec.

scute

.scute

Voronoi seeds → cellular loss

A Voronoi-cell image codec. The image is reconstructed from seed-defined territories: every output pixel belongs to its nearest encoded seed.

Corruption moves seeds, buckles borders, flares cell colours, and lets surviving territories annex space after truncation. Record order is intentionally not meaningful, so shuffle is not part of the format grammar.

Open the cell codec.

vermis

.vermis

Hilbert thread → flowing loss

A Hilbert-thread image codec. The image is sampled along a Hilbert curve as DPCM colour deltas, then repainted as one continuous filament instead of a square pixel grid.

Corruption accumulates downstream. A damaged delta becomes a travelling stain, bruise, or phase shift through the body of the worm.

Open the thread codec.

urumizuri

.urumizuri

Wet state snapshot → tectonic shear

A space-allocated layout codec native to the bakezuri performance station. Rather than calculating flat pixels, it maps raw environment grids—storing live ink load paths, local wetness maps, and paper-setting field variables as uncompressed byte records.

Because record sizing relies on rigid channel strides, corruption breaks base alignments. Manual byte editing triggers tectonic spatial cascades, causing the active ink layers to shift sideways and shear along geographic fault lines instead of throwing unhandled parser crashes.

Open the wet snapshot station.