AI art-generating programs are progressively getting better and better at creating hyper-realistic images. For the last few months, I have been using AI to create headshot images of completely fake people with the lighting style of a few famous photographers. In this social experiement, I wanted to see if the average person could tell the difference between images of real people photographed by real photographers verses imaginary people created entirely with artificial intelligence. Take my quiz and view the results in what I call "Real vs AI Headshots."
Every day, more and more hig- profile whistleblowers continue to share their dire concerns about how artificial intelligence is going to shape the future for humanity. We've all seen the interviews with people like Elon Musk, Demis Hassabis, Geoffrey Hinton, Sam Harris, and Mo Gawdat who consistently warn that AI-based programs like OpenAI's Chat GPT, Google's DeepMind, and IBM's Watson are only the tip of the iceberg in a race pitting computer science against human development. While these programs are no doubt leaders in the AI space, as a photographer, I'm much more interested in what artificial intelligence is doing to the creative world, specifically photography and videography.
Like many other photographers, I first became aware of AI art-generating programs back in the summer of 2022. Programs like Midjourney and DALL·E 2 were pretty impressive, but the results were still often cartoonish, overly mystical, and lacked the realism to seem like a real threat to the world of traditional photography. Over the last 12 months, these programs have gotten better and better at reproducing photorealistic images with both text-only prompts and image-to-image renderings. It is actually pretty comical to go back and rewatch any YouTube content from 2022 predicting where AI generation will be in a year and seeing where we have actually wound up. Comical might be too benign of a word to use; perhaps naïve or even shortsighted might be a better perspective on how fast this technology wound up progressing in just 12 short months.
So, how far along has AI-generated imagery come, and where does it stand today at the end of 2023? This was the question I sought to answer when I came up with my latest social experiment. Instead of asking Fstoppers readers, who are often full-time photographers, I wanted to see how my realistic but completely fake AI images of people were received by average, everyday people who aren't make a living with photography. Would they immediately recognize the soullessness behind my creations, or would they be fooled into believing they were infact real people captured by real photographers?
Take The Test For Yourself
Before explaining how these images were created or any of the specific results, if you want to take a digital version of this test, you can click the image below. The 44-image test consists of a mixture of real photographs taken by real photographers and AI images I created using the image-generating program Stable Diffusion. I've set a time limit of eight seconds per image to simulate the average glance most people would give any given photograph. Be aware though, as soon as you click the quiz button, the first image will appear and the timer will start.
How the Test Was Created
I never intended to create any sort of test pitting real photography against AI-created art simply because, at the time, I didn't think AI-generated images were good enough to fool anyone. It wasn't until I switched over from Midjourney to Stable Diffusion that I started to realize just how far along this technology had come. For me, Midjourney's images were simply way too stylized for anyone to find them real. The colors were too vibrant, people's skin was way too fake, faces were far too perfect with few imperfections, and overall, they looked more like something out of the movie Avatar than a print from a Richard Avadon medium format negative. Stable Diffusion changed all of that, though.
I'm not the most tech-savvy photographer, so installing everything need to get Stable Diffusion was pretty challenging. Unlike Midjourney or Open AI's DALL·E software that basically works the same for everyone, Stable Diffusion is based around open source programs and requires you to become familiar with software jargon like Python, Github, and special AI-training plugins called Loras, style models, and variational autoencoders (VAE). If that's not complicated enough, you can also install different graphical user interfaces to help make your workflow more visual instead of code-based. I wound up installing one called Comfy and primarily used a custom node or workspace called SeargeSDXL. I only mention this in the event you are as curious about Stable Diffusion as I was, because it's honestly not as intuitive and user-friendly as the aforementioned Midjourney and DALL·E AI-generating programs.
I spent the better part of a month tweaking and testing my workflow, but eventually, I was able to produce some truly spectacular AI-created headshots. All of my images were based around the photographic styles of two of the most famous headshot photographers, Peter Hurley and Martin Schoeller. Although I used word prompts only, because these two photographers have such a massive body of work online, the AI software was able to emulate their styles pretty easily. In order to get realistic renderings of human eyes, hair, and skin, I often had to employ different loRA models, upscalers, and refiner models. This took a ton of time, and honestly, even after I perfected a handful of images, I found myself constantly tweaking all these settings over and over again.
My goal was to generate over 100 usable AI-created headshots. Initially, I relied loRAs that other Stable Diffusion users had created that were trained to perfectly create celebrities, models, actors, and even porn stars (because this space is very heavily used to create NSFW imagery). After generating a handful of images that looked exactly like well-known people, I started scaling back those AI-trained models to just insert maybe 20% of someone's likeness in hopes of creating more generic-looking people. For my final test, I included a few of the celebrity doppelgangers because Peter Hurley and Martin Schoeller are both known for working with the real people in their own work.
For the real photographs, I simply mixed in a wide variety of actual photographs taken by real photographers. A few images come from Peter himself, and others come from his headshot teaching platform The Headshot Crew. The goal was to evenly mix real images with AI-generated images and keep the styling consistent with Peter and Martin's work only. Because both of these photographers are known for very stylized headshots, I wanted everyone who took my test to be limited to only seeing these types of headshots. In order to make sure every participant was fully aware of the photo aesthetic they were about to be tested on, I briefed each person on Peter and Martin's work first by letting them explore each of their websites.
The Medium
The easy way to run this test would be to simply create an online test or quiz and let each participant quickly click through a bunch of images and see the results. I've created that test above that you can run for yourself, but I wanted to make this experience more hands on. In order to make this a more intimate and visceral experience, I wound up printing all of these headshots through the photo-printing website Saal Digital. I've used Saal Digital in the past to create portfolio albums, posters, and photo art for my walls, so I knew the quality of their prints would give this test the immersive experience I wanted instead of simply handing someone a mouse or iPad and having them click through a screen.
Each image was printed as a 4x6 photograph on museum-grade FujiFilm Crystal Archive DP II Matte Paper. This paper not only brought all the colors and details of each image to life, but the matte paper also prevented fingerprints from ruining the experience for the next viewer. Even if only subconsciously, I think that seeing all of these images in a real, tangible print also made each participant feel more connected to the work. They became more real and valuable compared to just a JPEG on a computer screen.
I want to thank Saal Digital for sponsoring these prints for this test and being patient with me as I created this entire project. This took far longer than I ever could have imagined, but it was an experiment that I thought was extremely important to run and test. Regardless, if you are a professional photographer or just someone who enjoys physical prints of your own family and memories, I'd encourage you to continue printing your own photographs in a real, physical format. If you want to see all the print options Saal Digital offers, you can check out their website here. At the writing of this article, Saal Digital is offering Fstoppers readers a whopping 50% off all their products through this discount page.
The Results
Coming from a pre-medical background in college, I was super excited to not only run this photography-based experiment, but I was also curious to see if the hyperrealistic AI images I had created were good enough to fool the average person. The main question I wanted to test was: "are AI created photos using only word prompts good enough to fool the average person into believing they are real people?" Notice that I'm not as interested in seeing if professional photographers can tell the difference but rather the normal, everyday person.
My test sample in this video was fairly small with only about 20 people from my neighborhood. In the weeks before publishing this article, I have opened up my online digital quiz to dozens more people, and the results are pretty consistent. On average, most people are scoring about 40-60% when shown 44 images. In my test group of 6 elementary and high school age kids, the average jumped up to around 50-74%, but I'd like to get more participants to see if this increased perceptiveness holds true or if it's just an error based on a small sample size (and testing above average children).
I'm excited to see and post more results as this online quiz circulates around the internet. I'm prepared for it to be skewed a bit because of the format change and perhaps the difference in how digital images are displayed versus printed photographs. Also, since this quiz is originating from a professional photography website, I'm sure the results will be skewed a bit based on the occupation of many of the test-takers. However, I'm hoping in a few weeks time the number of people who participate in this quiz will be large enough to start building some strong opinions on how AI art generating programs are affecting people's perception of reality.
When I was a kid, I remember hearing a photograph is worth a thousand words. But today, with AI being able to create anything you can imagine in a truth-defying manner, we are inching closer and closer to a photograph being worth whatever the human mind wants to believe it to be worth. If AI-generated art can now blur or outright change the ability to determine truth and honesty, we might find ourselves living in a very scary and apocalyptical world sooner than we thought.
Scored 95% accurate.
https://take.quiz-maker.com/results309980-2963c961-TFY59OAT
Overall impressed with the quality of the AI photos. Wonky catch lights are a good tell.
It's interesting about the catchlights. Normal more daylight catchlights do look wonky and aren't always consistent, but with much of Peter and Martin's catch lights being triangular, square, or parallel strips, people who do not know anything about photography often think they are the fake ones and something doesn't sit right.
"We all know how stupid the average person is. Now realize that, by definition, 50% of the population is dumber than that." --Ivan Stang
I think about this concept a lot, especially with the right to vote. On one hand, I like the idea that everyone should have a say in our elected politicans, but on the other hand, I see the craziness on both sides of the American political system and I think "how are these people allowed to vote?" Then thinking about your Stang comment, imagine the people who get to vote that we don't even see on the crazy spectrum! I say all of this thinking I'm probably only in the top 30% of informed people and most times I don't feel educated enough to cast such an important vote myself.
AI is going to make all of this so much more insane. I really think humanity is going to lose our minds over this mis information, fake imagery, body dysphoria, fake wonderlust, etc etc. I full heartedly believe "instagram" is a massive problem for the human psyche and there is no doubt that AI will make it even worse. I hope I'm wrong about everything I just typed.
Cool quiz. I only got 28/44, despite using a relatively large, high-DPI screen. I am disappointed in myself.
One gripe with this article: the website (quiz-maker) asks users to provide their e-mail addresses before showing the results, without revealing what the e-mail address will be used for. I assume, quiz-maker will sell it to spammers at the next opportune moment.
There may be better platforms out there to host such a quiz.
Tom, i share this fear, which is why I used a trash email.
I've been taking photos of human people for a long time and thought I would achieve a high hit rate. I could only get a score of 23/44. This terrifies me and I share Patrick's view, this is taking on apocalyptic proportions. I hope that a mandatory labeling requirement will prevail. (Sorry translated with a translation program, I hope it is understandable)
Quiz maker isn't making an email list for themselves, I asked it to create an email list for Fstoppers. I could go either way on asking for an email or not but I chose to turn it on in the event we ever want to contact people who took the quiz for something. That said, as Klaus stated below, you can use a fake email and still get the results.
I tried soooo many quiz platforms and none of them would allow me to do what I wanted it to do. I needed single images shown at a time, ideally with a timer, the ability to randomize the images, it shouldn't make vertical and horizontal images show drastically larger and smaller (they were all resized to be close to the same size), I wanted a results page with custom text on results, a backend where I could see tons of statistics, and a way to integrate the final emails into Mailchimp.
None of the platforms I tried could do all of this and some of the ones that did were like $1500 a year and I only do a few quizes a year. Quiz-maker was also pretty expensive but I plan to end the subscription in a month or two. Making the quiz was probably the most frustrating part of this whole video concept.
Where Lee makes review of overpriced products Patrick makes review of AI software? Smart way of doing it, but wasn't Fstoppers kind of DPreview community? What is happening guys?
I didn't really review any software here. Stable Diffusion is free, open source, and you can do whatever you want with it. As for the DPreview community, we still have the community but Lee and I have always focused on making entertaining, educational, review videos in general. To me this video isn't much different than the Stun Gun Photoshoot or my video Makeup vs Photoshop.
Did I miss Patrick's score? Can't find it.
My score? I score perfect every time because I made the test!
Yep, you are right. Could be interesting to see your score in a year or two without reviewing anything.
8 seconds was a little short so I missed the first few. Then I got tired of so many. I quit.
wow.
What is this test good for? The results of the so-called AI are only a snapshot and may turn out to be completely different (more realistic) tomorrow. So what is the purpose of such a test? AI in image generation is like all other technological developments: They are here to stay and are constantly getting better. Until perhaps something even better comes along ...
Every photographer who develops and edits their own digital image files uses AI. The result is no longer a photo but a digital image that often has little to do with the original shot. In the near future, AI will be able to generate the complete image perfectly straight away without the need for an expensive camera, lens and all the necessary accessories. Great. Let's wait a little longer and the image-generating AI will be able to create perfect images. Every AI shot is a bull's eye!
My thought was we photographers know what AI is capable of creating, but when I look at my wife's social media feeds or other non photographer friends who share really bad AI avatars that look like cartoons, I started to wonder if the general public has any idea just how far along AI has come?
This video isn't really for photographers as most of us are already using this technology or are at least aware of it. This video is for the general public. If normal everyday people cannot tell what is real vs AI right now, then it's going to be impossible to know what is real down the future. I suspect every platform will have some sort of AI fact checking scanner much like Twitter or Facebook have but then we are all going to complain about the bias those checkers have. I personally think this is the most important dilemma humanity faces in the near future. Not global warming, or clean water, or population decrease, or nuclear war....I think it's the inability to know what is real and fake and our rather privative minds not being able to keep up with the technological evolution heading our way.
The test must come to the same conclusion as the test of a chess master against a chess computer. It is always a question of WHEN the new technology will be better than humans and not IF. As we are already at an advanced stage of this comparison, the test is simply obsolete. It's better to focus your work on the question of how we humans will deal with this, i.e. where humanity will develop when nothing can be regarded as real any more. Answers can be found in many science fiction films.
I worry about AI images of historical events and people I'm already seeing cropping up. At this point, I think I can readily identify them by identifying anachronistic styles of lighting, clothing articles, or pose for the supposed time period. But I find even now most people readily accept AI images as genuine historical photographs.
Scored 82% thank god
https://www.quiz-maker.com/results318902-6b0BaC0D-TFY59OAT
The only way I was really able to tell is when catchlights didn't match, or an eye looked like it was unnaturally drooping or too symmetrical.
Interesting quiz but I can't but ask: what's the point of creating AI headshots? What client would want that and for what purpose?
I can imagine using AI for example for creating mood boards with particular environment and colours, or for stock images, but headshots?
This experiment isn't about making fake AI headshots for your clients. Instead, it's testing whether humans can distinguish between the two right now at this AI infancy stage. I filmed other tests with these people with other genres of photography so I'll release those soon.
I have a few worries in general about this.
1) If advertising agencies and direct brands can now make AI human faces look so real that we can't tell the difference, that is going to put a lot of photographers out of business. It's easy to imagine all the types of images that could be created but in my example of headshots, this could easily be beauty shots used for cosmetics and skin products. Any advertising that needs a photo of a person could use AI to replace the model, the photographer, the hair and makeup artist, the retoucher, and possibly even the artistic director.
2) We very well could reach a point where the average headshot client can simply upload a bunch of photos or a deep lidar scan of their own face and have AI produce dozens of amazing looking headshots. I know it seems far fetched that these would ever look authentic but I think they will. I've seen software that is already doing this with expressions and 3D movement and it looks incredible (and it's still very very early stages). Check out this AI avatar creator using face mesh. If you were event remotely impressed with my avatar at the end of this video created by HeyGen, Ziva is going to blow that away once it becomes more accessible. https://zivadynamics.com/ziva-face-trainer
I know everyone wants to believe that we all want "real" photos and for certain things we probably do (weddings, family portraits, baby photos, ect) but if you are casting someone for a movie and lets say every movie now renders some part of the actors with AI because it's cheaper or produces a better result, maybe actor's headshots being made with AI isn't going to be a big deal. I can't tell what the future will bring but to not concern yourself with it at all and pass it off as not likely I think it a mistake.
Thank you for the thoughtful reply! I have to say my initial thoughts were quite wrong. When I saw a headshot photograph, I have always thought about a profile picture, a newspaper article (to accompany an interview for example), documentary, etc. Always the person being the subject. There’s no point in AI in these use cases because we photograph THE person, not A person. I never thought about adverts for skin products and similar, with a generic human in the frame.
Ziva is awesome (thanks for the link). Generating or significantly altering actors has been used in cinematography for some time, obvious examples of AI can be younger face of Harrison Ford in Indiana Jones, or of course Carrie Fisher in last episodes of Star Wars. In the more general use it’s ridiculed as deep fakes, but it does exist for a while now and deep fakes can be uncomfortably convincing. It’s always faking a particular human, not creating a generic one, however.
I’m cautiously excited about the future of AI in cinematography. Animated films have existed pretty much as long as films with actors, and they have been improving over time. From Walt Disney’s work with parallax all the way to realistic modern 3D. AI is the next step in making it more interesting and engaging, which is great news for viewers.
Yet, people often (at least in my country in Europe) go to the cinema to see the actor rather than the movie. And commercials touting “Starring Johnny Depp as …” suggest that this is quite universal. Side actors though…?
I just answered totally at random, since 8 seconds per image is not how I expected the quiz to be. But cool, yeah, they look pretty realistic at a glance.
8 seconds is to replicate the average time someone looks at a billboard or advertisement in a magazine or subway. It could be 5 or it could be 15 but the point wasn't to give people unlimited time so they can determine 100%. If you weren't even told you were being quizzed, you probably wouldn't even suspect AI to begin with.
Misleading title: should read "AI photos are almost indistinguishable from heavily digitally edited photos".
Both contain images where every colored pixel is altered, 1 just began with a camera.
I used mostly Peter Hurley's images in the test and he really doesn't do a lot of editing on his photos. He requires a makeup artists for his female headshots and sometimes his male headshots too. He might remove a pimple or something that is only there for that particular shoot but he doesn't go in and remove tons of skin and do deep frequency separation or anything like that. If you are referring to the images being digital and tweaked in Lightroom, yeah that's every single photo you ever see in advertisement ever. I don't think it's fair to dismiss modern day photography because it's been retouched. It's always been retouched and as I said, the examples I used aren't retouched much at all if at all.
I failed miserably in the quizz. But its a victory of sorts for the AI Technology 😏. That's perhaps the whole the point the quizz intended to highlight!