One face, any scene
A platform for a client whose name stays under NDA. It trains a model on a handful of photos of one person and keeps that face consistent wherever it goes: an astronaut helmet, a racing car, a business portrait, a cartoon, even a cat.
A creator builds the character once and reuses it for posts, images and short clips instead of starting over each time.
Lead developer, from scratch
I led development of the MVP over four months and wrote it full-stack. A backend developer worked under me, and the two of us designed the architecture from zero.
The models came from a separate AI team. A good part of my job was syncing with them: agreeing on what the product should be, what their side delivers and how the two halves talk to each other.
Fast around slow AI jobs
Training a personal model and generating images take minutes, not milliseconds, yet the product had to feel instant. Background services do the heavy image work, and the interface never waits on them: it shows training and generation progress, keeps a history of every result filterable by person and background, and tracks credits.
Access runs on credits that renew monthly, with a Pro tier for early access and unlimited generations.
Stack
