How to Keep AI Headshots Consistent Across a Whole Team
Aragon AI is built for one person at a time. Using it across 30 people exposes a different problem: every set looks fine alone and wrong together. Different crops, different background greys, two people lit like they are in different rooms.
This is the operations side — how to run a company-wide batch so the directory page looks like one shoot.
Why do individually good headshots look bad together?
Because each person picks their own style. One goes for a warm outdoor blur, another for a bright white studio, a third for a dark grey corporate drop. All three are defensible choices. Side by side on a grid, they read as chaos.
Consistency is a decision made once by one owner, then enforced for everyone. It is not something the tool can infer.
What goes in a style brief?
Write one page, send it before anyone uploads a photo:
- Background — name the exact style from the library, plus a hex-approximate grey for trimming.
- Crop — head and shoulders, eyes at roughly the upper third line, no hands.
- Framing — 1:1 for the directory, 4:5 for the website team page; everybody delivers both.
- Expression — one neutral and one smile, no laughing.
- Clothing — one to two levels of formality allowed, no logos, no busy patterns.
- File naming —
firstname-lastname-directory.jpgand-web.jpg.
That brief removes 90 percent of the back-and-forth.
How do you handle bulk purchases and billing?
Aragon AI sells individual packs, so a 30-person rollout means 30 purchases unless you are on a business plan with volume seats. Two ways to keep it manageable:
- Company card, one coordinator. The coordinator buys the packs in a single session so invoices stay together.
- Reimbursement, one invoice template. Cheaper administratively, worse for consistency, because people buy different tiers.
Check the current business plan terms for the exact seat minimum before you promise a number to finance. In practice, the break-even sits somewhere around 20 to 25 people — below that, individual packs plus a strict brief is cheaper.
What is the QA checklist before publishing?
Go through every image in one sitting, on one monitor, with all 30 open in a grid:
- Background grey matches within a few shades of the approved reference.
- Collar and shoulders sit at the same height relative to the frame edge.
- No stray text or logo on clothing.
- Teeth and ears checked at 100 percent.
- Glasses wearers checked for reflection artefacts.
- Colours corrected to match skin tone across the grid, not per image.
Reject rather than retouch. Regenerating a set costs nothing extra on a pack that already includes 100 images.
Where the process breaks
People upload old photos. Someone sends a company portrait from 2019 with a different hairstyle and the result is a person who is half-current. Set a rule: photos from the last 24 months only.
Remote teams mix lighting. Half the team shoots near a window, half under a ceiling tube light, and the generated skin tones drift. Ask everyone for the same setup — window, no flash, plain wall — in the brief, and offer to reject and re-shoot.
Contractors. Some will not want their face in a dataset. Do not make it mandatory.
Who this approach suits
Companies between 10 and 200 people with a directory page they care about. Recruiting-heavy teams where candidate experience matters. Anyone replacing a bad batch of selfies on an about page.
It does not suit teams where each person has a personal brand and wants a personal look — a marketing agency of creators will resist a uniform brief, and they are right to.
Quick Summary
- Company-wide consistency comes from a one-page style brief: background, crop, expression, clothing, file names.
- Batch purchases mean one pack per person; business plans usually beat individual packs above roughly 20–25 seats.
- QA the whole set in one grid session — matching grey, matching collar height, 100% zoom on teeth, ears and glasses.
- Reject and regenerate rather than retouch; surplus images are already in the pack.
- Only accept source photos from the last 24 months.
