AI Headshot Generator on Martini

Headshots are a hidden onboarding cost. An HR director welcoming 30 new hires at a fintech needs uniform, on-brand portraits for the team page, the company directory, the badges, and the slide deck the founder uses for fundraising. Booking 30 individual studio sessions runs into thousands of dollars and weeks of scheduling — and even then, three different photographers produce three different lighting setups, three different crops, and three different backdrops. The team page ends up looking like a collage of unrelated headshots rather than a single, professional roster.

The other half of the problem is mid-cycle hires. The team page goes live in March; two new partners join in May; the next funding round announcement happens in June. Now you need three new headshots that match the original session — same backdrop, same lighting direction, same crop discipline, same brand color script — but the original photographer is booked, the studio is unavailable, and the budget for a one-off shoot is hard to defend. Without a way to lock the look once and re-run it per new hire, the team page drifts into mismatch territory within a single year.

And the final break is platform-specific output. LinkedIn wants a square crop with the face well-centered; Glassdoor and the company website want different aspect ratios; the slide deck wants a transparent-background cutout for compositing onto a brand backdrop. Generic AI portrait tools deliver one square image per generation, no aspect-ratio variations, no cutouts, no consistency anchor for the next hire — leaving the marketing team to manually crop, mask, and color-correct every output before it ships.

Martini treats the team-headshot project as a multi-anchor canvas. Drop the brand-color backdrop reference as one labeled image node, the lighting-direction reference as another, and each new hire's submitted selfie as its own subject node. Wire all three into Imagen 4, Nano Banana 2, or Midjourney for the first hero pass and into Flux Kontext for subject-faithful refinement. Every hire's portrait inherits the same backdrop and lighting from the locked anchors, so the team-page roster reads as one cohesive shoot rather than thirty disconnected uploads.

Multi-model fanout for the first hire and template scale-out for the rest. For the founder portrait that headlines the press release, fan out across Imagen 4 (photoreal), Nano Banana 2 (subject fidelity), Midjourney (editorial polish), and Flux (high detail) on the same reference; pick the take that holds the brand best. For the remaining 29 hires, lock that winning model and the backdrop+lighting anchors as a canvas template — each new hire becomes a single new subject node added to the proven chain. The expensive iteration happens once on the founder, the catalog of hires scales on the template.

Downstream chaining handles delivery. Once each headshot lands, chain into the background-removal tool node for the transparent cutout the design team needs for the press kit, into the image-upscale tool for the high-res master used in the all-hands deck, and into a square-crop variant for LinkedIn while the original 4:5 portrait ships to the team page. The same canvas produces every aspect ratio and every delivery format from a single approved take, so the HR team never re-cuts headshots manually for each platform.