Why don't E-FITs look like actual people?

e-fit images of eight people

Everyone's seen 'E-FITs' of criminal suspects on the news and in the papers. Grainy and greyscale, or leering unsettlingly with improbable features, it's sometimes hard to see how the suspect could look like the composite and also be a real human person.

For instance, all eight of the nightmare-inducing faces in the main picture above are real e-fits issued by UK police forces, from 'Magic Tardis Hat' (top-right) to 'Rainbow Voldemort'(bottom-right).

Yet at the same time, videogame characters and customizable avatars are better than ever, with some being genuinely recognizable and others a little too close to the uncanny valley: counterfeit-looking humans that give you the creeps with realism rather than rubbishness.

So how come we can make a footballer look pretty damn accurate in Fifa, or perfectly recreate our (idealised) selves in The Sims, but some criminal portraits still look like a comic-book character that was photocopied 50 times?

Feature presentation

E-FITs have a long history. Originally, trained police sketch artists would hold a detailed interview with a victim who'd got a good glimpse of the suspect, in order to establish their features well enough to make a likeness. This technique is still used today (the two faces below were released by West Mercia Police relatively recently), but understandably, brilliant artists with sensitive interviewing skills are somewhat at a premium, so other methods have developed alongside theirs.

police sketch of two suspects

Traditional police sketches are still in use with some forces. Credit: West Mercia Police

In the 1950s, Identikit was created to help US police forces who didn't have a trained artist available, and involved laying clear strips with creepy disembodied features on top of each other until you made the right face. The Photo-FIT system, based on a similar premise and created by the UK authorities in the 1970s, introduced the 'FIT' name. Then came E-FIT, or 'electronic facial identification technique,' which – before the age of laptops – was impressively created on a portable PC (although it needed a big old tube TV to show the output). 

Image of Dolch computer

A Dolch 286-Pack portable computer used to create E-FIT images is now in the Centre for Computing History in Cambridge, England. Credit: Centre for Computing History

Like Identikit, the early E-FIT systems relied on the premise that someone who'd seen a criminal could accurately remember and describe their individual features, which could then be combined into the right visage. Unfortunately, human beings (with the exception of a small group) don't work that way, and tend to take in faces as a whole. 

In videogames, on the other hand, the artist (or you, if you're creating an avatar) has a visual reference right in front of them. They can see exactly how wide the person's nose is in relation to their mouth, and how deep-set their eyes are, and how all the features are spaced. This makes a huge difference to accuracy: if you perfectly remember all the features of the robber you saw, but get the spacing wrong, it could look like an entirely different person.

Take these examples, from Warwickshire and Dundee Police in the UK respectively, who have some implausibly-sized features:

e-fit images of two men

Police wouldn't have had much trouble catching these two if they looked anything like their E-FITs. Credit: Warwickshire Police, Dundee Police

Even if you're creating a digital version of yourself without a photo reference, you can play about with the nose size slider and the skin tone adjuster until it looks 'about right' (yeah, those shredded abs are pretty close to my doughy stomach, that'll do). So why isn't this the case with E-FITs?

Actually, it kind of is. The E-FIT system has been improved and built on by multiple modern approaches, the key one being E-FITV. Originating at the University of Kent in the UK, and now fully owned by VisionMetric Ltd, this system works much better with the way we see and remember faces – holistically, and there's a sort of nose slider, too.

image of police officer using a computer

A Sussex police officer using CD-FIT, a short-lived competitor to E-FIT, in 1993. Credit: Sussex Police

Facing facts

Instead of combining features one by one, E-FITV starts with a few details about the criminal (age, race, that kind of thing) and generates a set of faces based on them. Then, the witness keeps picking the face that's most like the person they saw, and the system refines and refines until it has a pretty good likeness.

For instance, you might pick face 2 of 9, at which point the system will generate 8 more faces with deviations from that one (a larger nose on one, a different skin tone on another) for you to choose from, until you look at the result and say "that's him". 

So what happens if the system isn't generating the face you need? Well, you pick the best fit, then get in there and mess with the details yourself. You can move and scale features, rotate them, add details like wrinkles, and even automatically adjust the age of the face. And if there are two faces you want to smush together, you can do that too.

The result is a full-color, detailed, realistic (ish) likeness that's more akin to a videogame character than a bad cartoon. Here are three modern E-FITs recently released by London's Metropolitan Police:

modern e-fit images of three men

On first glance, modern E-FITs almost look like photos. Credit: Metropolitan Police

So why don't we see images this good everywhere? Well, firstly, newer systems including E-FITV and the University of Stirling's EvoFIT are proprietary and have a pretty captive market, so they're not cheap. They also require training, which is something most UK police forces are already short on time and budget for. And even with the benefit of training, not everyone's going to be great at operating the software, so some faces will be better than others.

Secondly, we're just not using E-FITs as much as we used to, because tech's intervened in another way. We don't need to rely on fallible human memory if we've got a video of the suspect from CCTV or citizen journalists. So with those stretched budgets, it makes even less sense to invest in new face-making software over, say, a new camera that could get actual photos of the suspect.

That said, there are many police forces, both in the UK and other countries that have made the investment in modern face compositing packages, and that means they should be able to catch more criminals – but we'll have to say goodbye to a rogue's gallery of comedy composites.

TechRadar's Next Up series is brought to you in association with Honor

Latest in Tech
Ray-Ban smart glasses with the Cpperni logo, an LED array, and a MacBook Air with M4 next to ecah other.
ICYMI: the week's 7 biggest tech stories from Twitter's massive outage to iRobot's impressive new Roombas
A triptych image featuring the Sennheiser HD 505, Apple iPad Air 11-inch (2025), and Apple MacBook Air 15-inch (M4).
5 unmissable tech reviews of the week: why the MacBook Air (M4) should be your next laptop and the best sounding OLED TV ever
Apple iPhone 16e
Which affordable phone wins the mid-range race: the iPhone 16e, Nothing 3a, or Samsung Galaxy A56? Our latest podcast tells all
The Apple MacBook Air next to the Dyson Supersonic R and new AMD GPU
ICYMI: the week's 7 biggest tech stories from the best tech at MWC to Apple's new iPads and MacBooks
A triptych image featuring the Bose Solo Soundbar 2, Nothing Phone 3a Pro and the Panasonic Lumix S1R II.
5 trailblazing tech reviews of the week: Nothing's stylish, affordable flagship and why you should buy AMD's new graphics card over Nvidia's
The best tech of MWC 2025 examples, including the Nothing Phone 3a Pro, the Nubia Flip 2, and the Lenovo Solar PC
Best of MWC 2025: the 10 top tech launches we tried on the show floor
Latest in News
Metroid Prime 4
I reckon the Nintendo Switch 2 could launch with Metroid Prime 4 – here’s why
Samsung Galaxy Z Fold 6
New rumors predict a foldable iPhone will launch next year – and cost almost twice as much as the iPhone 16 Pro Max
Pebble smartwatch countdown
Pebble confirms its smartwatch announcement is just hours away
Logo of YouTube Shorts
Is YouTube auto-playing Shorts when you open the app? Well, you’re not alone - here’s how to fix it
Google DeepMind panel discussion
“More sovereignty and protection” - Google goes all-in on UK AI with data residency, upskilling projects, and startup investments
Nintendo Switch 2
Nintendo Switch 2 expected to have AI upscaling and I can't wait to finally play Tears of the Kingdom with upgraded graphics