Photographer faithfully recreates his work using text-to-image generator

You may know Antti Karppinen for his amazing conceptual photos and photo manipulations. He’s been merging illusion with reality in his work for a while now, and now he’s added an interesting twist to it: artificial intelligence.

Thanks to tools like DALL-E and Midjourney, you can create an image from pretty much anything you can spell. Antti was curious whether it will be possible with his work. Since it’s dreamy and surreal, Midjourney seemed like a logical choice – and it did a remarkably good job!

Antti tells DIYP that Midjourney doesn’t have similar features when creating new content as Stable Diffusion. However, you can feed it images as input reference and then use the prompt to guide it to generate something similar. Here’s the basics of how you do it:

To recreate his work, Antti placed one image at a time as an input and typed its content into AI. Midjourney returned a grid of four images, and Antti would select the best one and upscale it. For example, he used his image and a prompt “Dolores from Westworld holding a gun, sunset canyons” for this one:

“I’ve been playing with AI for almost a year now and I’m starting to get the hang of it even though the technology is moving at the speed of light,” Antti writes. “Learn a new way of doing something and the next day there is something new popping up that draws my attention.”

“It takes something like 30 seconds to one minute for Midjourney to create those images,” Antti tells DIYP. And it’s both amazing and a bit concerning how quickly the technology changes. “From images to animations generated by text prompts, this industry is changing many things, and also disrupting industries,” Antti concludes. “How we create images is not the same anymore.”

Take a look at more photos below and let Antti know which ones you prefer. The real ones or the AI-recreated ones? Make sure to check out more of his work (without AI touch) on his website, Instagram, and Facebook.