Any project- be it a short film, an ad, a teaser, a fashion campaign, or a UGC video-starts with a character.
A model. A face. But building that face?
It used to take forever.
Endless iterations… or tools like NanoBanana just replicating references- accurate, but not original.
So I built a system on @FLORA
Input:
– 3 image references (eyes, nose, lips)
– A single prompt for structure, skin, and mood
Output:
– a fully cohesive, hyper-realistic face
Now, instead of generating random identities,
You design them- feature by feature.
Make him walk?👌 Run? Laugh? Cry?🫶 Swap the hat? Talk?🫡
I built one character for an open-source Mac AI app. It shows what each AI agent is doing (thinking, writing code, searching), and new moods, outfits, and even species can be added without rebuilding it.
51 animations. 3 characters. 1 file. Built in Rive, so it lives inside the app and reacts to what the app is doing.