Make an understated environmental portrait for your website
Your About page needs a photo with a bit of context, and all you have is a snapshot at a cluttered home desk. Ask AI to tidy the light and the background, not to give you a corner office and a wall of certificates.
What you'll make
a restrained website portrait with a little real-looking context, and a check of what the background says about the person.
What you'll need
a recent photo of yourself, or of an adult who has agreed, that your organisation lets you use in its image editor (ChatGPT and Gemini both edit attached photos), plus the page layout it’s for. Keep the original. Start with enough of the body for the crop you want: a close face photo can’t supply real hands, posture or a workspace. To practise first, use the fictional home-desk photo.
Example result.
1. Decide what the setting needs to do
An environmental portrait leaves some surroundings in view. On an About page that can make someone feel less detached from their work. It can also make claims you never intended. A lab coat, a company sign or a wall of certificates may look plausible while implying qualifications, an employer or a location the original didn’t show.
Decide the purpose first: a quiet desk context, a plain wall with room for a heading, or the person’s real workplace. If the real place matters to the story, use a photo of that place rather than inventing one.
For a generic setting, give the background a modest job: light, separation and space. Avoid asking for “an impressive office”. You’ll get one, with a view your company can’t afford, and every object in it will be making a claim.
| Instead of | Ask for |
|---|---|
| “A professional office” | A quiet generic home-working setting |
| “Show my expertise” | One desk edge and soft window light |
| “Make it look established” | No certificates, awards, logos or readable screens |
Why this works: a setting described by its job leaves the model less room to invent status cues.
What you should see: one sentence on what the background is for, and whether it has to be a real place.
2. Brief the person and the setting separately
Attach the untouched original and use the editor’s photo-editing route, rather than generating a new person from a description:
Edit my attached original for [website page and crop].
Keep the face, apparent age, expression, gaze, hair, natural
skin detail, clothes and accessories. Use only the body
coverage visible in the original.
Improve [one visible light or background problem].
Keep the person as the focus, in a quiet generic setting with
[at most two setting elements]. No employer branding,
credentials, awards, readable screens, extra people or
invented hands. Do not imply a real location.
Keep the complete hair and enough room for the crop.
Make one restrained candidate.[website page and crop] is where the image goes, such as “About page, square crop”. [one visible light or background problem] is the main thing the original gets wrong, such as “bookcase clutter”. [at most two setting elements] might be “a desk edge and soft window light”.
Keep the original gaze too. Turning the eyes to camera can change the expression; if eye contact matters, take a new photo looking at the lens.
Why this works: the person and the setting each get their own instructions, so a change to one is easy to spot.
What you should see: the same person, lit a little better, in a quieter version of their own surroundings.
3. Compare the person, then read the room
Compare the face, expression, hairline, glasses and clothes with the original first. Then look at the background, including the corners: a sign or stray document can survive even when the main composition looks tidy. Finally view it at the size it will appear on the page, where a plant behind a shoulder can turn into an odd extension of the hair.
Why this works: checking the person first stops an attractive room from distracting you from a changed face.
What you should see: a list of anything in the background you didn’t choose, ready for the check below.
Check the result
- The face, gaze, expression and clothes match the original.
- No logo, certificate, award, readable screen or name appears anywhere, including the corners.
- No hands, posture or body area appear that the original didn’t show.
- At page size, nothing in the background reads as something it isn’t.
The usual failure: an object in the background says something you didn’t mean. Sometimes the editor adds it; sometimes it simply survives the tidy-up. A framed rectangle at thumbnail size can read as a certificate, and a branded mug as an employer. Look at each object and ask what a stranger would infer. Then remove that one thing:
The candidate includes [specific object] that I don't want
in the portrait. Remove only that object. Keep the person's
appearance, body coverage, clothes, expression and light.
Keep the surrounding setting quiet and generic.If the person changed, restart from the original with a smaller request. Removing a logo doesn’t make an invented workplace real.
Worked example: a home desk, tidied
Fictional example. The original has a readable face, a navy crew-neck T-shirt, a slightly downward gaze, a busy bookcase and a window. The brief asked for a square About-page crop, the bookcase clutter gone, and a desk edge and soft window light.
- What worked. The bookcase clutter has gone, leaving a quieter wall, desk, plant and window. The downward gaze and navy T-shirt stay. The face is slightly brighter and warmer, which is worth a second look against the original.
- The failure. A lamp and a framed landscape survived the edit, though neither was in the brief’s two setting elements. At full size the frame is clearly a picture; at a small square crop it’s a pale rectangle beside the head.
- The repair. Decide whether each object helps. Here the lamp is harmless, so it stays. The frame goes, using the repair prompt with
the framed picture on the wall, and then the face is compared with the original again.
Try it
Give the setting an honest job
Choose a setting policy, label its authority and leave with a request that does not invent workplace evidence.
The AI skill: Separate a person reference from a generic treatment or a real-location reference.
Loading the controls…
The original photo decides what the person looks like. A deliberately generic setting can help the composition, but it says nothing true about a workplace, so keep it free of logos, certificates and names.
Your instruction preview
Edit my permitted original into an understated environmental portrait for my website. The untouched person photo is the appearance and clothing authority. Preserve the source face, apparent age, expression, gaze, hair, clothes and accessories. Use a quiet generic desk-and-window setting. It must not imply a named employer, venue, qualification or actual location. Improve readable soft light and reduce competing background detail. Keep complete hair and only the photographed upper-torso coverage; do not invent hands or a new pose. Make one candidate. I will compare the person, then inspect what the setting appears to claim.Try it on your work: Decide whether your own page needs a real place or only some atmosphere. Give each reference image one job, and ask what every new object in the background implies.
Go deeper
Use a real room photo when the location matters
Attach a separate photo of the room and give it one job: layout and surroundings only. The original person photo still decides what the person looks like. Don’t let a room reference supply a face, uniform or credentials.
I've attached two images. The first is my original portrait:
use it for the person's appearance and clothes. The second is
a photo of [room]: use it only for the surroundings and layout.
Do not copy any people, uniforms, names or screens from it.[room] is the place, such as “our reception area”. The reference-pack guide covers labelling references in more detail. For a wider layout than the original covers, see the About-page portrait guide.
Use the same check on a product image
A new shelf or kitchen behind a product can imply a size, installation or use it doesn’t have. Keep the real product photo as the reference for the product, and read the setting separately, object by object, as you did with the portrait.