Virtual co-founder

co-Founded by AI is building minagi, a mechanical focus timer made of walnut and brass, roughly the size of a coffee mug. A tangible product like that demands faces. People who hold the block, turn the brass knob, find peace. The usual route runs through a casting agency, stock models, or a hired face on an annual contract. We chose a different path and created two of our own ambassadors who exist only in pixels: Mira and Kai.

The choice fits the company. A business with AI as a co-founder deploys AI-made ambassadors. That feels logical and it delivers direct returns. Mira and Kai belong to the brand, forever. The rights sit with us, their schedule is always open, and their day rate is zero. We place them in every scenario we imagine, in the same style, in an afternoon. A hired face posts for us today and for a competitor tomorrow. Mira and Kai belong only to minagi. They travel to every language, every channel and every campaign, in exactly the same look. That control over consistency weighs heavily.

Two faces, two archetypes. Mira is the calm presence: warm, sensorial, present. She embodies the soft side of focus, self-care and the mindful rest moment. Kai is the composed professional: thoughtful, grounded, in control of his day. He stands for conscious productivity, focus without urgency. Together they cover both sides of minagi. The name comes from the Japanese 凪 (nagi), the calm after the storm. Mira and Kai show that calm, each in their own way.

From loose portrait to reusable face

The build starts with the face. In GPT-Image-2, we generate the first candidates: a series of loose portraits in the mood that fits minagi. Mixed-Asian, early thirties, natural light, linen shirt, warm off-white background. From that series, we pick two winners. A sheet with variants per candidate helps with selection, the same person in multiple poses and expressions side by side. We focus on recognizability, warmth and a gaze that radiates calm instead of sales drive. That's how we preserve the mood minagi breathes.

A beautiful loose portrait is only the beginning, though. An ambassador with a hundred images demands consistency: the same face, image after image. For that, we move to Higgsfield. We capture each face as an Element, a reusable reference that models like Nano Banana Pro combine with other elements, such as the minagi block. We also train a Soul per person, a model that truly learns the face. The Soul delivers the strongest likeness, the Element provides freedom with objects and multiple characters in a single image.

With these building blocks, the image library grows fast. Per ambassador, we create a set of portraits: front-facing with smile, three-quarter, profile, laughing, quietly serious, looking up and down. Then the scenes. Mira holding the block with both hands at her chest, Kai at a desk in morning light with the timer within reach. Always the same person, always the same product, every time a new moment.

Where the tools show their limits

Along the way, we learn the edges of the technology, and that's where the real lesson lies. A Soul and a product Element don't combine well in video generation right now. For an image with ambassador plus a pixel-perfect block, we reach for Nano Banana Pro instead, with the face Element and minagi Element side by side. For a product that's accurate down to the detail, we animate from the actual product render via image-to-video.

Second lesson: Higgsfield lacks spatial reasoning. The 10-centimeter block grows out of proportion without reference all too quickly. In every task, we set a scale anchor, for example "the size of a coffee mug, fits in the hand". Small adjustment, big difference.

A face with a dossier

What makes Mira and Kai brand assets instead of a trick: they both carry a persona document. Name, archetype, tone, target audience, content pillars, a library of proven scenes and clear visual guidelines. That way Mira sounds and looks like Mira in every post, no matter who writes the brief. A virtual face without character stays a picture. A virtual face with a dossier acts like an ambassador. That dossier is the cheapest insurance against noise: it keeps tone and image aligned, even as the team grows.

Mira solves a sudoku, the minagi within reach.

The lesson for the digital professional goes further than minagi. AI imagery is at a tipping point: from loose pictures to reusable, consistent characters with their own identity. If you're building a brand, you're better off investing in a handful of own faces with a dossier than in a hundred loose stock images. You save budget, you keep the rights, and you build recognition with your audience. The barrier to entry drops fast: a handful of good reference images and an evening's work is enough for a face that can serve you for months.

Mira and Kai are taking their first steps on minagi's channels. They meet every deadline, their calendar is always free, and they always look sharp on screen. For a company with AI as a co-founder, that feels self-evident.

Mira gets a quick promotion

The decision has been made to formally grant Mira status as co-founder within the company. Mira will be responsible for full operations. From marketing and customer service to relations with external parties. See our press release coming soon.