# STEAL MITO: extended case study

[Project brief](https://www.arantxabarcia.com/agents/steal-mito.md)

These are first-party portfolio claims, not independent verification. Null or missing fields mean not published. A preview clip is not the complete film. Images and videos have not been transcribed. Reference works are influences, not projects or credits by Arantxa. Case-study extracts follow source text order and may retain presentation fragments; use the project brief for factual lookup. Editorial draft appendices are excluded.

## Project

[Source section](https://www.arantxabarcia.com/work/steal-mito#project)



## the heist

[Source section](https://www.arantxabarcia.com/work/steal-mito#case-process)

The heist

Steal the frame

STEAL MITO was made to launch the MITO shop, and the whole thing was written, generated, cut and graded in two days.

The idea came out of the problem. A shop launch usually asks you to photograph objects and hope somebody wants them. We inverted it: if the object is worth stealing, you do not have to argue that it is worth having. So the film is a heist, played straight — a vault, a vitrine, a laser grid, a man in a balaclava — and the thing under the glass is a cap.

It is also a joke about the brand. The line on the cap reads MYTH MAKING IN PROGRESS, and the invitation is to steal the frame rather than buy the merch. Lupin would approve. Briefly.

Worth saying plainly, because the images do not show it: at the end of 2025 the models were nowhere near as precise as they are now. Faces drifted between shots, hands came out wrong, a cap would change its shape from one generation to the next. What now takes one pass took five, and the work was less about writing prompts than about repairing what came back.

## how it was built

[Source section](https://www.arantxabarcia.com/work/steal-mito#case-method)

The method

How it was built

The seventy per cent happens before anything is generated. Coherence is an architecture of references, not the model's memory — everything you do not define in advance, the machine invents. For a two-day job that is not a philosophy, it is the only way to finish.

Cast

Actor first: validate frontal portraits, then reuse the face and fixed wardrobe verbatim in every prompt. Build scenes only after this, or identity and clothing change between shots.

Refinement

Refine with image references: change the eyes or match a hat, keeping instructions in the filenames. Target only what is wrong, preserving what works instead of rewriting the whole prompt.

Space

Then the room, built and locked before anyone walks into it. One space, one light logic, one lens: everything that follows has to belong to this corridor or it does not go in.

Object

Then the object, treated as the thing the whole film exists to protect. It is lit like jewellery and framed like evidence, so the heist is legible before a single laser appears.

Action

Only at the end the action, because by then nothing is left to invent. Character, wardrobe, space and object are already fixed, so a shot is just a decision about where to put the camera.

The take

The payoff. Two days from the first portrait to the grade.

## the object

[Source section](https://www.arantxabarcia.com/work/steal-mito#case-object)

The product

The object

The cap is real, and it is the only thing in the film that is. Everything built around it — the vault, the lasers, the man — exists to make one embroidered line worth crossing a room for.

The cap

Front

Side

Back

MYTH MAKING IN PROGRESS, embroidered on the side. The film never says the line out loud; it just puts it under glass and makes somebody come for it.

Steal it from the shop