CARLO OPPERMANN

AI Journal

Part portfolio, part time machine.

DominosHomo CorporatensisDino CookiesStylizing FootageAlways something new

AI in 2026

AI is now everywhere. And without a doubt its going to change everything we know about media. The trajectory is a bit more obvious now with Seedance 2.0 and Seedance 2.5, we are clearly not far off from productions being completely AI and nobody caring. As I write this, I am getting my scripts ready for all the shortfilms I never made.

Anxious and excited

I am feeling both - massively anxious and massively excited. Generative AI will change our media landscape, culture and society fundamentally and probably not only for the better. There are many things to be critial of.

However: As a filmmaker with a lot of ideas but limited resources it can be a dream come true and add many, many possibitlies to the toolbox. So for now, curiousity wins.

What you see here is part portfolio, part time machine. Scroll down and you’ll find my AI experiments scattered across the years as I have been tinkering around with every tool I got my hands on since Dall-E 2 in 2022.

Dominos

behind the scenes

The failures are just as fun as the successes. For every finished film, there are at least ten times as many broken, bizarre, or downright cursed generations. Honestly, half the joy is in watching AI trip over itself—it’s its own genre of comedy.

Homo Corporatensis

experimental AI short

Character Consistency

This was an idea I’d been carrying around for years—something involving mammoths, prehistoric office workers, and other things my budget stubbornly refused to cover. Then AI image generation finally got good enough to keep characters looking mostly consistent from shot to shot.

Dino Cookies

animated with KlingAI

Img 2 Video

Think Jurassic Park, but with dough. When Kling 1.6 came out, I made real photographs in my kitchen and turned them into a tiny short where cookies suddenly sprang to life.

Stylizing Footage

testing in january of 2024

Remember That Look?

Before proper video generation showed up but img2img was available, we hacked our way through with image-sequences.

The result? Chaotic flickering, faces mutating every other frame.

It quickly had a strangely iconic look of its own. And this was just the start.

Always something new

early experiments, 2022–2023

AI feels like a firehose of shiny new toys—every week there’s another “revolutionary” tool, and half the time I’m just poking at them without any plan except not wanting to miss out. It’s chaotic, it’s overwhelming, its exhausting at times…

But it’s fun to look back.

The earliest experiments in 2022/2023 could barely pass as fluid video. Yet the potential was undeniable.