Measurable synthetic creators
Build consistent virtual creators for product content, campaigns and owned social accounts — with every asset scored against a calibrated identity threshold before it is approved.
A pilot is one creator, one calibrated identity, and five scored product scenes — enough to see the threshold hold on a creator built for your category.
One approved identity, held across every scene, framing and lighting condition — and checked rather than assumed. Still the same person in post four hundred.
A declared voice per creator and one never-say list for all of them. Captions are checked against it before they are approved for publishing, not after.
The creator holds, wears or uses the product in a scene you specify, still clearing the same threshold their approved set cleared on day one.
Every frame passes a scoring gate on the way in. A frame that scores below the creator's calibrated threshold does not become an asset — the score, the verdict, the reason and the cost are written to a ledger beside the image, so an approval or a rejection can be re-examined months later. How the threshold is set →
What it is for
The point of a creator who stays the same person is commercial. An audience follows a face, and a brand pays for reach against that face. Consistency earns recognition. Transparency protects trust.
The cast
Each one declared, seeded and rendered through the same pipeline — the same code path, the same gate, the same ledger. Different ages, heights, builds and presentations. The same identity standard applies to every creator.
One of them has been taken all the way through approval and calibration; the rest are at the candidate stage, which is where a creator sits before a dedicated cohort run.
The evidence
Nine faces prove range. Range is the easy half. Below is one creator across different scenes, framing and lighting — and what makes that believable is not that it looks right to you. Every frame cleared a threshold set by measuring her against 90 strangers.
Measurements from the running system, read when this page was built — on one creator, who is the one that has been calibrated. A second creator needs her own cohort run. Full methodology and signal inventory →
How it works
Age, height, build, colouring, presentation. Some traits are immutable and some are free to vary by scene — and which tier a trait sits in belongs to the trait, not to whoever is asking.
Candidates are rendered and scored, worst first. You approve faces, not filenames. The approved set becomes the reference every later frame is conditioned on and compared against.
The threshold is not a number somebody liked. It is set by scoring her against a cohort of people who are not her, then against her own approved frames, and reading where the two stop overlapping.
Frames under the threshold never become assets, and the reason is written to the ledger next to the cost.
What you get
Not a services menu — these are the things the studio produces and hands back, every one of them carrying the record that makes it checkable.
Traits fixed where they must not move and free where they may, seeded so the same person can be rendered again tomorrow.
The frames you accepted, one at a time — the reference the creator is held to from then on.
The number that decides what passes, handed over with the error rates it produced rather than as a setting somebody liked.
Shot to brief with the product in frame, each one carrying the score that let it through.
The voice itself, and the record of which captions cleared it and which were turned back.
Every decision the system made, still legible months later: what was kept, what was turned back, and what it cost.
Where it is today
Both columns are on the page on purpose. A tool that measures its own output does not get to be vague about its own state. The detail, with numbers →
Start with one creator, calibrate that identity against a stranger cohort, and see the score on every scene that comes back.
Accounts are created by hand. Send your details and somebody will follow up.
Every creator is transparently synthetic — protecting audience trust, brand safety and platform compliance.