Why AI Photos Sometimes Don't Look Like You: 5 Quality Checks for Events

When an AI portrait doesn't look like the guest, it is usually because the model changed facial proportions, hair or distinctive features while applying a style, or stretched a half-length photo into a composition it could not support. Events can judge quality on five checks: likeness, feature retention, wardrobe, composition and brand safety, and should test before the day.

Why do AI portraits sometimes miss the likeness?

An AI photo booth is an interactive on-site installation that captures guests in the moment, generates stylised images with an AI model and delivers them on the spot. During generation, the model is balancing two jobs: keeping the guest recognisable and applying a new style. The two pull against each other. The stronger the style, with bigger changes to wardrobe, setting or lighting, the more likely the model is to alter the face too. How an AI Photo Booth Works walks through each step of the process.

The usual causes fall into four groups:

• The style changes defining features. If a theme asks for different hair, a new hair colour, closed eyes or sunglasses, guests will not recognise themselves.
• The composition goes beyond what was captured. The booth captures guests from the waist up. A theme set as full-length, seated or reclining forces the model to invent a body, and proportions suffer.
• The style was designed around one kind of guest. Themes with fixed evening gowns or set hairstyles give uneven results across a mixed audience.
• Capture conditions. Lighting, angle, glare on glasses or a partly covered face all affect how accurately the model reads a face.

Most of these causes can be spotted in a pre-event test, which is why testing with a realistic mix of people matters more than reviewing a few polished samples.

What are the five quality checks?

The table sets out the five checks, what to look at and how to test each before the event.

Checks 1 and 2: how do you judge faces and features?

Likeness is what guests care about most. Don't judge a theme on its single best sample. Test it on several people of different genders, ages and skin tones, and see whether it holds up for everyone. Feature retention matters just as much: hairstyle, hair colour, glasses and open eyes are how people recognise themselves. A theme that keeps changing those details is wrong for a corporate event, however striking the image.

When choosing themes, favour styles that express themselves through light, backdrop and colour over styles that rework appearance. 6 AI Portrait Styles for Corporate Events covers the styles corporate events tend to use.

Checks 3 and 4: wardrobe and composition

Corporate guests are mixed, and so is what they wear. If a theme generates new outfits, make sure both men's and women's versions are appropriate, and offer an original-outfit option for guests who would rather keep their own look. For composition, test at the real booth distance and angle rather than with phone selfies, because the framing is different. Also think about how guests will actually stand at the booth. Guests in groups, guests holding drinks or guests wearing hats all change what the camera sees, so include a few of these situations in your test.

Check 5: how is brand safety protected?

For brand events, appropriateness comes first. FaceVibe Booth's Campaign Custom lists excluded elements at brief stage and shares samples for the organiser to approve before the event (AI themes). When reviewing samples, look for third-party brands, magazine titles, stray text or anything unsuitable for the occasion. How to Brief a Custom AI Photo Booth Theme explains how to write exclusions into a brief. For events with press or senior guests in the room, agree in advance who reviews the samples and who has the final say. A clear approver avoids last-minute debate about whether a theme is appropriate.

How should pre-event testing work?

FaceVibe Booth tests the confirmed themes before the event, checking AI portrait quality, generation time and the download flow (How it works). Organisers can also book a session at our Causeway Bay office to try the booth and see results for themselves. When testing:

• Invite colleagues of different genders, ages and hairstyles, including people who wear glasses.
• If you expect group photos at the event, include them in the test.
• Review every image against the five checks and note any theme that is inconsistent.
• Test both looks, AI outfit and original outfit.
• Confirm the waiting time per guest so it fits the event programme.

Keep the test images together with your notes. If a theme needs a tweak, they show the team exactly what to change, and they give your approver a clear record of what was signed off.

What reduces the problem on the day?

The most effective design choice is to let guests choose. FaceVibe Booth guests pick their favourite from several AI versions. At BusinessFocus Awards 2026, the event offered two looks, AI outfit and original outfit, with three candidate portraits generated for each; guests chose the one they liked best and downloaded it by QR code (Case studies). Several candidates plus an original-outfit option means there is almost always a portrait guests recognise and want to keep. It also helps to let guests know up front that they will be able to choose from several versions. Knowing there is a choice makes people more relaxed in front of the camera, which tends to produce better portraits.

On-site staff help too: they guide guests and assist at the booth, for example suggesting the right spot and angle, which cuts errors caused by capture conditions. Theme details are on the AI themes page. If you also need to explain to colleagues how guest photos are handled, see AI Photo Booth Privacy: 8 Questions.

Finally, keep an eye on the first guests of the evening. If a theme behaves differently from the test, the organiser's contact can flag it to the team early.

FAQs