ChatGPT's New Skills Resident Evil 4 Remake Galaxy A54 5G Hands-On TikTok CEO Testifies Huawei's New Folding Phone How to Use Google's AI Chatbot Airlines and Family Seating Weigh Yourself Accurately
Want CNET to notify you of price drops and the latest stories?
No, thank you

Police-sketch software puts faces on fiction characters

New Tumblr blog applies facial-composite program to descriptions of characters in a variety of novels--from Dr. Robert Vaughn in J.G. Ballard's "Crash" to Humbert Humbert of "Lolita."

Brian Joseph Davis

What if your favorite sci-fi or fantasy character broke loose from the book you were reading and went on a rampage?

Your first step (after scrambling under the bed) might be to call the police. And they, of course, would want the suspect's description--to hand off to their sketch artist.

That's where Brian Joseph Davis comes in.

In a mashup of high and low culture, the writer and artist has been creating police composites based on descriptions of characters in novels: Dr. Robert Vaughn from J.G. Ballard's "Crash," Gary from Colson Whitehead's "Zone One," Aomame from Haruki Murakami's "1Q84," even Humbert Humbert from Nabokov's "Lolita" and Edward Rochester from Charlotte Bronte's "Jane Eyre."

The unconventional portraits form the basis of Davis' Tumblr blog/crowd-sourced art project "The Composites."

Davis uses a contemporary take on Smith & Wesson's '60s-era "Identi-Kit," a collection of mix-and-match cards showing various types of facial features and hair styles. As many readers probably know, investigators used the tool while questioning witnesses, to create an image of a suspect. Faces is a higher-tech, software-based version of the same approach.

Davis interprets the written descriptions of characters, then chooses facial elements from Faces ID. Thus, Dashiell Hammett's classic limning of "Maltese Falcon" protagonist Sam Spade:

Samuel Spade's jaw was long and bony, his chin a jutting V under the more flexible V of his mouth. His nostrils curved back to make another, smaller, V. His yellow-grey eyes were horizontal. The V motif was picked up again by thickish brows rising outward from twin creases above a hooked nose, and his pale brown hair grew down--from high flat temples--in a point on his forehead. He looked rather pleasantly like a blond Satan.

yields the following facial composite:

Brian Joseph Davis

Hammett's word-portrait is relatively specific, but that's not always the case. As Davis told The Atlantic's Megan Garber, one thing he's been made very aware of while working on the project is the shift in literary style that's taken place from the 19th into the 20th and 21st centuries--lengthy, detailed character portraits are harder to come across than they were a hundred or so years ago.

In fact, in the crowd-sourcing aspect of the project, Davis solicits requests for characters/subjects from readers of the blog, and he's let it be known that the protagonist of "The Catcher in the Rye" isn't likely to be showing up anytime soon. In a "suggestions update" on the site, he writes: "Unfortunately, there will be no Holden Caulfield. At a glance, the entirety of his self description amounts to 'I have a crew cut.'"

In cases that lie somewhere between Hammett's hard-boiled detective and Salinger's searching teen, Davis the would-be police artist does some sleuthing and tries to find clues to appearance in other aspects of the text, or in the era that produced the work in question.

"What I found really interesting was that Sam Spade ended up looking so '1930s,' pretty close to a sinister Dick Powell, just based on Hammett's descriptions," he told Crave in an e-mail. "Writers do draw on their era--films, theater--and their era's famous faces or types, and I think always have. To go back to 'Crash,' it predated punk a few years, but there's a line about "self-cut hair" that does lend itself to presenting Vaughn as a punk prototype."

We all form images of characters while we read, and it's a common experience to be disappointed by a film's depiction of a beloved literary figure. We asked Davis if he'd been caught off guard by the way any of the characters had turned out.

"My wife is a novelist and we tried out the program with one of her characters first," he says. "She had a blast, and was a little surprised by the character she had actually described."

Which goes to show that language is a tricky, subjective thing. We suggested that it might be interesting to take the project a step further and ask viewers of the blog to use Faces ID to post their own renderings of a given character, with everyone interpreting the same descriptive passage. That seemed to jibe with Davis' vision of the crowd-sourced aspect of the blog.

"I'd love for that happen," he wrote. "It could be a new kind of folk art."

"I made 'The Composites' explicitly for Tumblr.... And really, it was Tumblr's mass, distributed curatorial character that I wanted to investigate," Davis wrote. "I posted this only a week ago and thought it would be a slow burn with some community involvement, maybe. I certainly didn't expect it to be this popular this fast."

So what accounts for the popularity? And why haven't people shied away, given the filmic disappointments mentioned earlier?

"I've seen a couple of commenters saying that this is fascinating but they felt it was a kind of a corruption of the imagination..." Davis wrote. "I don't see that. In fact I'd say this highlights the power of reading and imagination. I've had a lot of requests for characters that have seen big film adaptations, and it's not just fandom--I think it's because the readers suggesting them know that even a live-action adaptation doesn't quite capture what's in a writer's words and in the readers heads."

In fact, there's an educational version of the Faces software, which its makers say can be used with younger children for "enhancing drama and reading appreciation by allowing students to develop images of historic or fictional characters."

In any case, it's not as if these renderings are super-realistic. Part of the weird appeal may well have to do with the strangeness of the facial composite itself.

"With this particular style, I think the not-quite-3D rendering makes for an uncanny effect," Davis says, "especially on literary characters, who don't exist--and even rendered like this, still don't quite exist."