Good visuals are still gated by budget
Every business now needs more images than it can afford to make. A catalogue needs a shot per variant. A campaign needs a frame per channel, per format, per language. A founder pitching on Monday needs a hero image on Sunday night. The demand curve for visual content went vertical; the supply side is still a studio, a photographer, a retoucher and a two-week turnaround.
The result is a quiet inequality. Companies with creative budgets look credible. Everyone else uses stock that three competitors are also using, or ships nothing. We think that gap is a tooling problem, not a talent problem — and generative models are the first technology capable of closing it properly.
What we are actually building
Metabob is a pipeline, not a prompt box. Today we run open diffusion weights behind our own layer: our style families, our controls, a refinement pass for detail at high resolution, and a safety screen on every generation — and we are training our own models to sit underneath it. That matters because control is what separates a novelty from a tool. If you cannot lock a brand's colour, hold a composition across a set, or re-run a frame with one element changed, you cannot use the output for real work.
So the roadmap is deliberately unglamorous: fewer party tricks, more of the boring controls that make a generated image shippable — seeds, negative prompts, aspect discipline, in-place editing, per-customer fine-tunes that stay isolated, and an export that is genuinely print-ready.
How we work with customers
We incorporated in 2026 and we onboard in batches, so every new team gets a real conversation rather than a signup form. Design partners bring live briefs — a catalogue, a campaign, a deck — and what they run into shapes what we build next. That loop is the reason the controls are the way they are.
