Welcome back to the AI For Humans Newsletter!

So…will AI ever be boring again?

I started to wonder that sometime last week after ChatGPT 4o’s Image Gen broke the internet with Ghibli-core everything. There used to be full down weeks where not a lot would happen but now? Not so much.

The Ghibli-fication of the Internet has taken AI fully mainstream in a way we haven’t seen before and just yesterday OpenAI announced that everyone now has access to it, including all free users.

The Ghibli memes are fun and, as we mentioned in the last newsletter, pretty controversial but they’re also overshadowing just how wildly powerful this model is. And while OpenAI has quietly pulled back on what it allows, what’s still possible is… kinda magical.

So let’s talk about what you can actually do with it.

Make YouTube Thumbnails (or Anything) From a Drawing

Draw something extremely rough, upload it, and ask ChatGPT 4o to "turn this into a YouTube thumbnail." It knows what to do — colors, layout, text placement, the whole thing.

But you can also replace your face on other, more well-known thumbnails which Youtuber Samir Chaudry showed by placing his face on Mr. Beast thumbs.

Simple Four Panel Comics

The model "gets" comics. Just ask for a simple 4-panel strip about “a cat who wants to be president” or “an AI that discovers it has feelings” and the prompt doesn’t even have to be that complicated.

Unhinged Teenage Diary Collage Ramblings

We love the X creator fofr who comes up with all sorts of creative uses of image & video models. They made a format that looks like a chaotic 2003 high school diary: magazine clippings, gel pen handwriting, weird spirals, and emotional overshares.

Fictional Character Resumes

Want to know what Darth Vader puts under “Skills”? Or how Napoleon sells his “team leadership”? Or Ash Williams’ “Objective”?

Prompt 4o to “make a resume for [character] including fake job experience, skills, and a short cover letter.” It nails the tone. You will laugh.

Time Travel Via Parties (Or Really Anything)

4o Image Gen has very good baked-in control over where your subjects stand in the frame allowing you to keep them steady while you move them through time. This example from Cyberstrike on X gives us parties in four different decades.

It’s not just that this tech is impressive (it is). It’s that it’s now in the hands of hundreds of millions of people — and they’re all discovering totally new ways to use it every day.

We’re just at the start of this wave.

More on Thursday!

-Gavin (and Kevin)

We 💛 This - Runway Gen-4

I mean to we had to give the new model a try, right?

Like most model launches, it doesn’t come day 1 with many of the features we appreciated from the previous Gen-3 model. It’s missing things like camera control, restyled first frame or… speed. But we imagine those will come with time.

Despite the relatively sparse controls, you can get pretty far simply with text prompts due to it’s excellent coherence.