Meanwhile, in a Parallel Universe, I've Lead a Very Interesting Life - Deep Faking Yourself With EverArt AI

Could I be staring in Luc Besson's next Sci-fi epic?

Technically this isn't really Deep Faking but 'Training an AI Model on Yourself' isn't quite as eye catching a headline.

Deep faking is the act of replacing one person's face (and sometimes voice) in an image or video with another. It's often done in fantasy casting for films where the face of one or more actors is replaced by another to get an idea of what that actor may have been like in the role.

One particularly famous deep fake is Robert Downey Jr and Tom Holland replacing Christopher Lloyd and Michael J. Fox in the school hallway scene in Back to the future.

But I digress... Deep Faking yourself into situations that never happened is now one of my new favorite things to do with AI Image Generators. I don't know why I hadn't thought to do this much sooner as it totally feeds into my somewhat suppressed narcissism.

EverArt

I got to trial EverArt.ai, a new text to image AI generator totally focused on training image models for your needs. You get 25 free generations before you need to pay for a monthly account.

The models (called 'Styles' on EverArt) are very easy to train. I simply uploaded 24 images of myself where my face could clearly be seen.

From there the site works like most text to image AI generators. Type what you want to see with as much detail as you can, then hit the generate button. EverArt allows you to choose one or more of your selected styles, and allows you to pose transfer from other images too. It can also make suggestions for prompts - which can help when you're developing your own prompts.

Strangely you can only generate four images each time with no way to change the number of images. Which means you only get 24 free images with one left over from the trial you can't use.

I used all my generations on inserting myself into parallel universes where I'm way more 'windswept and interesting' (to quote Billy Connolly). 

Not every image was a banger but the success rate was pretty high in terms of not getting anything that look straight up weird or deformed. Here's a run through the highlights. Click the images to see larger versions.

The Life Where I'm Way More Successful

AI Prompt: A sophisticated businessman, in a gray tailored
suit, commanding attention amidst the hustle and bustle of
a vibrant city street. His posture exudes confidence and
authority, with one hand casually resting in his pocket.
The textured fabric of his suit catches the light, adding a
touch of elegance. The scene is bathed in warm golden
sunlight, casting long shadows and infusing the
atmosphere with a sense of vitality. The surrounding
buildings, adorned with sleek glass facades and colorful
signs, reflect the dynamic energy of urban life.

The first thing I wanted to try was a recreation of myself as my, usually, cartoon avatar. Essentially me in a grey suit. I did type in my own prompt but then saw a suggested prompt that was way better than mine so I went with that.

Honestly all four images generated are so good I'd almost swear I actually went on a professional photoshoot to have them taken. I may even consider going the full vest and jacket style suit, instead of just the jacket I usually have.

I don't think I've ever looked this sophisticated.

The Life Where I'm a 1940's Action Hero

AI Prompt: A 1940s action hero escaping from a
deadly pursuit through the jungle with confidence.
He looks amazing.

I was picturing an Indiana Jones type look in my head but what came out was more of a version of me as a tropical jungle war hero.

As you can see from my simple prompt only one image really depicts me 'escaping' from anything, however, I can't complain. In the main image my arms have never looked that rugged and amazing!

The Life Where I'm a Nineties Rock Star and Then Try to Make a Comeback in the Late 2000s

AI Prompt: Promotional photo of a nineties male rock star
performing live on stage. Behind him is his band
members rocking hard. [Pose transfer image of Shaun
MacGowan, bottom right.]

I actually generated these in reverse order but showing them in this sequence makes for a better narrative.

As a nineties rock star I have a mullet that's almost as good as the one I actually had in the nineties.

This was the only generation where I attempted to use EverArt's pose transfer feature, so I've included the image of The Pogues front man, Shaun MacGowan, that I used so you can see how well the AI did.

The only problem with these images was the faces on any other person in the shot. It looks like the AI struggles a little in deciding whether the background people should look like me or not? What you end up with is some very disturbing, derp faces.

AI Prompt: Promotional photo of a male rock star
performing live on stage to a full audience. Behind
him is his band members rocking hard.

The comeback images from the late 2000s really went off the rails with faces, including mine. There was really only one good image, with the rest having, what I could only describe as an audience full of zombies!

Both prompts for these images were my own wording. It was how bad the comeback images looked that prompted me to change the description to a nineties rock star and to also use pose transfer.

I wanted to look a little more rock'n'roll and less like a Dire Straits tribute band!

The Life Where I Survive the Apocalypse and Wander the Wasteland with My Trusty Blue Heeler(s)

AI Prompt: a confident, rugged, male survivor
in an apocalyptic, urban wasteland city, surrounded
by fantastic, rusting, scavenged vehicles. He wears
a long, open, worn out coat, heavy boots, and has
various weapons in his possession. At his side is
his trusty Australian, Blue Heeler, sheep dog.
The sun is setting casting a warm light.

After the successful businessman set this was quite possibly the coolest set of images I generated. While it was my own prompt, it definitely shows that the more detail you include the more likely it is you'll get an exceptional image.

It's pretty obvious I was going for the Mad Max aesthetic, and I wasn't disappointed.

I would've like to have seen at least one, more creatively constructed, looking vehicle in the background but at least the old rusted trucks made sense. Would've been weird to see a Prius (like they'd even survive an apocalypse).

Interesting that the AI really took to heart the word 'blue', not just giving me a Blue Heeler dog, but also matching scarf, and even a nice blue coat in one image.

The Life Where I'm an Action Movie Actor Staring in a Sci Fi Blockbuster

AI Prompt: A dynamic, high-contrast movie character
poster. The subject, a fierce and determined hero, stands
tall in the center, commanding attention. Radiant light
filters through epic clouds, casting a dramatic glow.
The hero wears an intricately detailed suit, shimmering
with metallic accents and glowing energy. Surrounding
the hero, a futuristic cityscape shimmers with neon lights
and towering skyscrapers. In the background, a vast,
starry sky stretches into the horizon. The overall mood
is intense and heroic, capturing the essence of epic
adventure and thrilling action.

Since I only had one more shot at generating four images I thought, wouldn't it be cool to see a character poster of me staring in a Science Fiction, action movie.

For my prompt I only entered the words 'A movie character poster' then I let EverArt build out a detailed prompt from that with its suggestion feature.

I don't really know what I was expecting but the main image here kind of came full circle from my businessman images, and is oozing, futuristic, John Wick vibes!

Certainly a lot cooler than the three alternate images where I'm wearing some weird kind of body armor - or that long coat with a tie that makes me look like I escaped from Zack Snyder's Rebel Moon film.

---o ---o--- o---

I don't think I'll be signing up to EverArt for a paid account anytime soon. The pricing is just a little too steep for the amount of casual use I'd get from it.

Since the Creator and Pro accounts have a 'buy more images as needed' option I'd love to see a free or significantly cheaper basic option that lets you buy 25 images at a time, as needed.

Aside from that I very much enjoyed creating these images and would have loved a chance to see how this AI would go creating images trained on my cat art (probably not well but you never know).

My overall take away though is, I should invest in a properly fitted, grey, business suit, and start walking around in it like a boss. It's definitely a good look for me!