The desk deals you up to four low-fi sketch directions — drawn from the client’s brand brain, at the post’s real shape, in about 45 seconds. Pick one, push back, or mix. No design literacy required.
Sketch directions scale by plan — 2 on Solo · 3 on Studio · 4 on Agency
Grounded before the first pixel: every direction is concepted from the client’s Visual Creative Brief — our read on their brand, distilled from their knowledge base and refreshed whenever it changes. The agency picks what goes to press — full quality happens after you choose, never before.
Streams in at the post’s real aspect ratio · ~45 secondsThe Visual Creative Brief distils the client’s knowledge base — their customers’ worlds, their product in scene, the mood and light that fit — and every image direction is concepted from it. It regenerates only when the brand brain changes.
A vision-equipped AI designer looks at the image with you, asks what to change, and edits the actual pixels — every step kept in a version tree, so nothing is ever lost. Refine chips come from the client’s real brand assets.
An AI art-director scores finals against the brief it wrote. Broken hands or garbled text are an automatic reject. The “★ Best” badge only appears at 90-plus — below 80, it plainly says so.
Two reference photos put a real person into generated scenes. We trusted it enough to run it on our own founder first — and shipped the result. One flat product photo does the same for what they sell.
Consent is part of the workflow, not a footnote.
Reference photos live inside the client’s workspace, with their permission on file — and they’re deletable any time. When a client leaves, their faces leave with them.
One flat photo renders the actual product — validated on worst-case, text-heavy luxury packaging: wordmark and medallion legible in four of four renders.
Van livery, coffee cups, signage — logos land with real lighting and perspective, routed lockup-vs-mark by the size of the surface they sit on.
When text inside a generated image is wrong, Re-Text finds it, classifies it, and cleanly rewrites or removes it — output at the exact source dimensions. No re-rolling the whole image.
Text generated inside an image is best-effort — that’s the state of the art, and it’s why Re-Text exists. Text on the template layer renders exactly, every time. We keep the two separate on purpose.
The mood board is style-only, never content — structurally: exemplars are distilled to text once; their pixels never reach the generator. And the AI doesn’t learn your taste behind your back — you promote what you like.
The darkroom light is on
Start free with 2,000 credits — paste a client’s URL, open the image desk, and react to sketches drawn from their world. Sketching is free; you pay at the finalize.
No card · No trial countdown · No per-client fees