An engine, built to turn stories into animation prompts.

Client

Personal project

Year

2026

Type

Desktop app

Role

Product Designer & Build

01 — Starting condition

Built to give story ideas a proper system.

Story and animation ideas move fast: a concept becomes reference images, then specific decisions about pacing, staging and camera. That process deserved a system of its own, one place to hold the story, the assets and the decisions together, with a clear path from that structure straight through to animation-ready output. So that's what got built, owned end to end rather than assembled from someone else's platform.

TBC

02 — Approach

Decide the rules before touching the interface.

Two decisions did most of the work. First: no server, no cloud database. Every project lives in its own folder, so it can be backed up, moved, or opened on another machine without anything else needed. Second: nothing changes without being written down first. Before a new feature touches how a project is saved, the change is documented and tested, so nothing breaks work already in progress. Small discipline, and it's exactly what makes the tool one you can trust.

TBC

TBC

TBC

03 — What was delivered

How the engine works.

Turns a written scene into a ready-to-use prompt for both a storyboard and the animation itself, checking automatically that everything lines up before it's sent off, saving real money along the way. A structure that's been rebuilt 13 times without ever losing a saved project. A proper Mac app, installed and used daily. Close to 250 automated tests, because a personal project earning trust still has to behave like one.

TBC

Schema

13 revisions since v1, with every entry point, opening a project, importing one, or running it from the command line, guarded against loading a version the app doesn't support.

Storage

One self-contained folder per project, holding the story, assets and moodboard together, reconciled automatically every time it saves.

Character references

A fixed contract per character (five main views, six palette swatches, five detail crops), so the model always knows exactly what to trust and why.

Distribution

Wrapped and installed as a real macOS app, no server involved.

Testing

246 tests across 32 files, with checks that fail the build if the output drifts by even a byte.

TBC

TBC

TBC

04 — Outcome

Parked the interface. Kept the engine.

The most interesting result was realising where the real value actually lived: not in the screen, but in what powered it. Most of the real thinking happens in conversation. So the interface was set aside as a simple viewer, and the logic behind it, the part that actually does the thinking, now runs on its own and plugs in wherever it's needed. Still actively developed.

Next Project