How AI-Generated Images Will Displace the Stock Photo Licensing Business

The media has manufactured a fear amongst us all that GANs – Generative Adversarial Networks – will bring the demise of democracy and civilization. (GANs are the AI behind deepfakes, in case you’re wondering). Sure GANs are a threat. But I’m adamant that they’re largely an opportunity. An opportunity for entrepreneurs.

If I were starting a GAN company today, I’d either be using it to create legal synthetic media like Synthesia or a stock imagery company.

Recently, I was inspired by a project called Generated Photos:

We are building the next generation of media through the power of AI. Copyrights, distribution rights, and infringement claims will soon be things of the past.

Generated Photos

Essentially they’ve trained a GAN to generate headshots of people that don’t exist, which they’re turning around and allowing anyone to use – royalty-free stock imagery. So far, they’ve created 100,000 faces, which in the grand scheme of things is miniscule. But it’s a start.

This work isn’t just a one-off novelty. We are building an entire toolset that will let anyone harness the creative power of AI as easily as integrating an API.

Generated Photos

Eventually, their platform will be a la carte, allowing people to pick and choose the inputs of the faces they want generated.

If you extend five or more years out, it’s not hard to imagine how this generated stock photography could displace the stock photography licensing business.

GANs, The New Getty 

Generated Photos has an emphasis on headshots. I’m anticipating another company to emerge soon with an emphasis on a different type of photo – architecture, nature, product shots, etc.

Not long from now, there will be a collection of companies like Generated Photos that mastered GAN-created images in their respective fields. Then someone will merge them.

Imagine being able to generate images with the same variety that you’ll find on a site like Getty, but with the selectability of a “build your own pizza”. It’s like having your own personal painter, illustrator, and photoshop artist on demand, creating any variation of image you want – brighter sky, fewer trees, and cardinals flying through. Bam! It’s made.

GANpaint, created in unison between MIT and IBM, uses a GAN to allow a user to paint or erase specific features of an image. They’re showing us that this vision may one day be possible on a grand scale.

Will this tool exist in ten years? Probably not. Or at least not at the quality we want. Will a GAN-based stock photo company put Unsplash or Getty Images out of business in the next decade?

Very unlikely.

Given the wide range of possible inputs and variables, this is a tough vision to execute. Hence, the reason Generated Photos started just with headshots.

But this “pie in the sky” type of idea is exactly the basis for the AI companies we’ll see emerge in the 2020s and have great impact by the end of that decade going into the 2030s.

GANs are a very promising form of AI in the entrepreneurial space. They need work. But I believe there’s a plethora of million-dollar GAN businesses and at least a couple billion-dollar ones waiting to be founded.