Mon Jan 08 - Written by: Brendan McNulty

Creating photo realistic AI Images

For fun and flattery

AI-generated portrait comparison showing different styles

The Experiment
I had this (somewhat lame, potentially awesome) idea for a presentation: a video of me hanging out with Elon Musk. Step 1? Creating still images of myself that I could use as a base.

My goal was simple (but surprisingly challenging): find a tool that could create high-quality, AI-generated images that looked like me in different styles and poses. Ideally, these images would look like I’d done a professional photo shoot—without actually having to book one.

Over a couple of months of iteration this turned out to be harder than I thought.

The Process and the Struggle
I started with MidJourney, a tool I’d seen hyped everywhere. I uploaded reference images of myself and experimented with prompts. At first, I was hopeful. Then… not so much.

  • They kind of looked like me—if you squinted.
  • The resemblance wasn’t strong enough to be instantly recognizable.
  • And the hands were definitely weird

Frustrated, I left Midjourney by the wayside. I remembered seeing some great AI-generated portraits on Pieter Levels’ Twitter (his demos were really good). He created PhotoAI, so I decided to give it a shot.

The PhotoAI Breakthrough
PhotoAI turned out to be a completely different experience. The pre-set photo packs—like Tinder, Glamour, and Work—were an interesting way of churning out a lot of shots. I uploaded a collection of photos of myself, making sure to include shots from all sides, good lighting, and plenty of angles. (Pro tip: err on the side of more pics rather than less.)

The resemblance was great.

  • The dropdowns and checkboxes in PhotoAI made the process easy to navigate—no complex prompts required.
  • The styles looked professional and, more importantly, like me.
  • PhotoAI even generated a few options with hair, which was a nice change since I’ve been bald for 20 years (even bald men need a little flattery)

It wasn’t perfect, but it was close.

Reflections on MidJourney and PhotoAI
Looking back, I can see how these tools are built for different types of users:

  • MidJourney is for the tinkerers and nerds—the people who love experimenting with prompts, iterating endlessly, and getting creative with styles.
  • PhotoAI feels more plug-and-play. If you want fast, accurate results, it’s hard to beat.

The Outcome
By the end of the experiment, I had:

  • A collection of AI-generated images that were eerily accurate (and at times flattering 😀)
  • New profile pics and marketing assets I could use for presentations and beyond
  • A fun, new way to experiment with AI

Pro Tips for Beginners:

  1. Prep Your Photos Carefully: Upload a variety of high-quality shots from different angles with good lighting. The more diverse your photos, the better the results.
  2. Iterate, Iterate, Iterate: If you’re using MidJourney, don’t settle for the first attempt. Experiment with different prompts and styles to refine your results.
  3. Go Beyond LinkedIn: These tools aren’t just for headshots. Use them to create marketing assets, profile pics, or just to have fun!

Want to Try It Yourself?

  • PhotoAI: Best for no-fuss, high-res images that look like you
  • MidJourney/Flux: Perfect for creative experimentation and pushing artistic boundaries
  • DALL-E: Worth a look, but less suited for high-quality portraits

Join me as I do 52 AI experiments in 2025. From teaching me to DJ to analysing interviews I’m documenting all I learn.