This week, GPT-4o was officially released and with it came a flood of images and a trend where every company under the sun started using the new agent for marketing. I swear, every brand that does RTM-style content had their own version featuring illustrated avatars of their presenters.

So yeah... the FOMO got to me. I was like, okay! I have to try it too.

I gave ChatGPT my favorite meme and asked it to turn it into a Ghibli-style image. I wrote, “Ghibli-fy this!” to my favorite meme of all time It gave me something… but not exactly in the style I had in mind, and I couldn’t figure out what I did wrong.

Unlike my previous article (deep research vs mobbin for ux exploration) thank god my partner was around to save the day. He told me: “You don’t do that directly in ChatGPT. You need to use Sora!”

So I logged into Sora which took a few steps, not gonna lie, because I hadn’t signed in before. It was a bit of a hassle, so just a heads-up if you’re new to it too. Once I got in, I saw this middle feed filled with creations from other users, which was honestly super fun to scroll through. What’s cool is that it doesn’t show the most liked ones, but rather the most recent — which is such a bold move, very different from DALL·E or Midjourney where not every image is a hit. With Sora, you get two options both stunning and super high quality. It even nails text in images (finally!), which I think alone deserves an award.

I’ll admit it took me a minute to figure out how to actually create a new image. I was looking for a button at the top left or right of the screen… but turns out, it’s at the bottom, just like a regular chat.

Another thing that tripped me up: all I wanted was a Ghibli-style image of my cats. I uploaded a photo, but then realized I had to switch the dropdown from “video” to “image.” That was confusing like, if I upload an image, it should know I want an image. And if I ask for a video, then give me video instructions, right?

Luckily, the system was overloaded at the time and video generation wasn’t available so I got an error that nudged me in the right direction and basically saved me from myself.

One more thing about the user experience: I don’t fully get why we have to choose an aspect ratio instead of just defaulting to the original size of the uploaded image… but honestly, that’s probably the trade-off for not having a clunky multi-step wizard. It’s a high-level UX move and honestly? Respect.

Here are a few of my favorite images I saw from other users:

P.S. I later found someone on LinkedIn who shared a full guide, and I highly recommend checking it out if you haven’t tried making an image yourself yet.

See you next time and thank you for reading!