"Imagine a world where a creative director walks over to a workstation, and says, 'We need a family—two parents, two kids—sitting down and playing a game together,'" says Daniel Jester, setting up an episode of The Creative Operations Podcast. "The person at the workstation punches a few keys, adjusts a few things, and—boom—out comes a unique image that's exactly what the creative director asked for. Not a mockup. Not a set of casting photos. The complete image is ready for use because you already own the license. You paid to license the technology, the algorithm that uses an ever-growing library of hundreds of thousands of images which have also been properly licensed, to generate this unique imagery."
For some of you, this description of stock photography on steroids might sound paradoxically efficient, while for others this marks a career-changing dystopia. However it feels, it's part of the future and even the present. That's why Mark Milstein, Chief Operating Officer of vAIsual, joins the pod to explain his company's technology to generate synthetic photography.
For the full conversation—it's a can't-miss—stream the episode on Amazon Music, Apple Podcasts, Spotify, or our website. But for highlights on Mark and Daniel's chat about synthetic photography, read on.
→ Want to learn more about what Creative Operations is? Read this definition article!
What's the Current Readiness of Synthetic Photography?
Mark's company, vAIsual, has what they're terming "the world's first commercially available algorithmic camera." Working from photos of real people, the technology generates "legally clean" datasets based on biometrics—in other words, generated people based on real people.
"At this very moment in time, we're able to generate real humans, on the fly, by typing into a command line, a search box, simple query: 'happy man with beard, smiling,' or 'unhappy man with beard, angry,' or 'young boy with red shirt and red hair in a neutral expression,'" Mark says.
What's next for this technology is combining these AI-driven models with backgrounds made from the same algorithm.
"Very shortly we'll be combining those humans with any variety of backgrounds," Mark says, "whether that be interiors or exteriors, landscapes—all kinds of a variety of environments. That will create synthetic photography—because our people don't exist—that rivals anything that comes out of today's DSLR cameras, with equal complexity."
Wait, So How Do You Generate These People?
It's not the same thing as rendering a character for a video game, Daniel reiterates because these aren't merely animations derived from an artist's depiction. These are biometric datasets combined into a person.
"Right now we have a studio, and we're shooting six days per week, and we have been for the past year and have committed approximately 300,000 images, which have been taken from well over 1,000 individual models, and those images are then fed into our algorithm," Mark says. "That algorithm then generates hundreds of new, never-existed-before, synthetic humans from that raw data."
So is There Anyone Who Could Claim That the Model is Based on Them?
Synthetic models come from real people "at a certain point," Mark says. "But what comes out of the algorithm doesn't resemble, in any way, shape, or form, that original person. There's no ability to even link one to the other."
With vAIsual's human models having no rights to the company's synthetic results, clients of this technology are spared from concerning themselves over rights, clearances, and usage of the images.
"All of our models have signed biometric releases," Mark says. "All of our images are GDPR-compliant. And so we have acquired those rights from these models, which then allows us to make legally clean results."
"No release forms," Mark goes on. "No rights restrictions. Full freedom from any legal hassles. You know, if I have a pharmaceutical campaign that's pushing a cream to possibly solve a sensitive problem that maybe some models might not want to put their face or name to, this allows for that."
And These Legally Clean Results are Called 'Synthetic Humans'?
"We like to refer to them as 'synths,'" Mark says.
"Hmm?" Daniel says.
"That's our go-to word at this moment," Mark elaborates. "Ultimately, somebody else might come up with a better word. But we call them 'synths,' as opposed to just 'reproductions' or 'outputs.'"
How Real Do 'Synths' Seem?
To Mark and his team, "synths" are lifelike enough to conjure feelings from company members.
"I'm laughing a little bit because there have been moments here in the office where we've actually felt like we wanted to develop relationships with some of the outputs we've made," Mark says. "Some of the people look so sympathetic—the kinds of people that we'd want to be friends with. We've said to our friends, 'Hey, look at this guy!' or 'Look at that girl!' We've said to ourselves, 'Wow, if they walked through the door right now, we'd be instantly drawn to them, in the most human sort of way.'"
So Then What Will Happen to Human Models?
For many in the industry, having no model release forms sounds nice. No paperwork. No legal nonsense to negotiate.
But no models? Are we ready for a future where we tell models, "Thanks, but we can take it from here"? And what about photographers and production assistants?
If vAIsual's technology went widespread, many of their skill sets would be repurposed and instead, you'd work with a "synthographer," as Mark calls it, a person whose skill set blends photography and graphic design.
Daniel can see the path for this hybrid role becoming more popular.
"With this technology, there's a path for people who have these skills and expertise in photography and lighting and composition to come in and up-skill themselves to become synthographers, and be able to generate these images," Daniel says.
"Right now, the synthographer does exist," Mark says, "just not with that title." It's someone who might traditionally work in Adobe's ecosystem, layering content to fashion never-before-seen content.
"And those skills that they presently use will just be better focused at creating synthetic photographers," Mark says. "We're just going to change their tools, right? Rather than using Adobe Premiere, Photoshop, After Effects, Illustrator, and so on, they're going to use the vAIsual tool over any of our competitors' tools to generate photography or a piece of content that integrates elements of synthetic photography."
So That's It? Most of Us are Either Future 'Synthographers' or About to be Unemployed?
Contrary to how some of this might sound at first hearing, this is not the end of the photo industry as we know, if you ask Mark.
"We don't believe that synthetic photography will wipe out the photo industry as it is presently," he says. "I think it'll be another means by which content creators or creatives within the advertising industry source their images. In other words, they have another choice. And I think another thing that'll happen is that really well-made, really highly creative photography will have greater value. There will be much greater value added to human-generated content when it's compared to its synthetic counterpart. Synthetic media, as I said, is more aimed toward a price-sensitive user from the very beginning. And so, therefore, higher-priced photography will continue to retain its value all throughout the entire content creation trip."
And…exhale.
"As we like to say on our website, and to investors and potential partners, throughout history, men and women have visually expressed themselves using every type of method and surface imaginable," Mark says. "Cave paintings. Oil paintings. Celluloid, glass, tin. Digital cameras. Your telephone. Every one of those tools has a limit. This is a completely different thing."
What's Next for Synthetic Photography, and Where Can We Try It?
Synthetic content from vAIsual is already available through German stock image providers PantherMedia and SmarterPics. For the company to see widespread adoption of their work, Mark believes they'll want to reach other traditional licensing platforms, like Getty and Shutterstock. The long-term plan is to make the technology available through an API, allowing clients to whitelabel the technology.
Want to hear more about synthetic photography, riffs on "college things," and a shoutout to Daniel's favorite regional chain for personal banking, stream the episode on Amazon Music, Apple Podcasts, Spotify, or our website.
