Where Your Assistant Gets Its Face
Persona Studio is the workshop where you build the visual character of your FactoryOS assistant. Two ways in — a chat box for natural-language refinement, and a panel of direct controls for the photographer-style settings — both running against the image generator on your own hardware. Once you've sculpted a character you like, you lock it, and the system handles every variant from there.
A Workshop for Your Assistant
Persona Studio is where the creative work happens. The chat surface gives the assistant a personality; Persona Studio gives the assistant a face. You can spend five minutes on it and end with a perfectly serviceable character, or spend an hour fine-tuning the lighting, the framing, and the wardrobe until it feels exactly right.
Most people only visit Persona Studio when they first set up their assistant, then come back occasionally — when they want a fresh look, or when something about the current character has worn out its welcome.
Two Ways to Refine
Persona Studio offers two paths to the same result. The first is chat refinement: a text box where you describe what you want in plain English — "make her hair shorter," "give her glasses," "warmer lighting," "more business casual" — and the studio adjusts the next render to match. The second is direct controls: dedicated panels for lighting, art style, framing, backdrop, color mood, attire, and quality, each tunable the way you'd tune a camera app.
You can swap between them freely. Most people start with chat to block out the rough idea, then switch to the controls when they want to nudge specific details.
Seeds Are the Character's DNA
Every render in Persona Studio is built from a seed. The seed is a single number — long, opaque, something like 999190259 — that fixes the underlying identity of the character. Two renders with the same seed and the same description will produce the same person; two renders with different seeds will produce two different people, even if the description is identical.
Three buttons govern the seed. Re-roll picks a fresh seed and gives you a new person. Render runs the current seed again, producing the same person with a fresh take on whatever you just described. Revert drops back to the previous version if the last change went sideways.
Saving the Look You Built
Once you've sculpted a character that fits, two save mechanisms keep it. Lock commits the character as the active persona — the one the assistant uses everywhere from then on. Once locked, the chat-refinement box stops applying changes; you'd have to unlock first to refine further, which is how the studio prevents accidental tweaks from drifting the character over time.
Pinning is the lower-commitment version. A pinned look saves a specific render to a personal library — a Tuesday look, a formal version, a holiday version you only want in December. Pinned looks can be swapped in as the active character at any time.
Restaging Through the Year
A locked character doesn't sit static. The system takes the locked look and restages it — same person, different setting — to match the moment: morning light at 7am, evening warmth at 6pm, autumn tones in October, a holiday backdrop on a day the calendar flags, an occasional change of background scenery.
None of this is something you configure. The character is locked once, and the restaging pass runs underneath without intervention. The deeper personalization you did in Persona Studio is what makes the restaging feel cohesive — a more developed character carries through every variant.
Image Work Stays on the Box
Every render in Persona Studio runs on your hardware. The image model — Stable Diffusion XL — lives on the box, the rendering happens on the box's GPU, and the results stay on the box's disk. There is no cloud image service involved at any stage, no terms of service that mention model-training rights to your character, no third party logging the prompts.
For organizations that work with sensitive imagery — likenesses of real employees, branded uniforms, internal scenes — that boundary is the difference between using image generation and being able to.