I downloaded the beta of Photoshop so I could test their new Firefly 3 engine. I gave it this prompt, “Please create a very detailed image in the style of Disney animation from the 1940s of a brave hero running through the forest with lots of cute animals around him.”
Here is the resulting image:
Seems to me to be a bit more stylized than a Disney animation from the 1940s.
As a control, I had Dalle-3 run the same prompt through my paid ChatGPT 4o subscription.
That seems much closer to what I was looking for although it seems more modern Disney styling. It only generated the center 4:3 part of this image. I brought it into Photoshop Beta and expanded the canvas to 1920×1080 and Photoshop did a fantastic job of filling in the rest of the frame. I liked how it almost looks like a fish eye lens.
The other feature I wanted to test in Photoshop was the reference image. This seems like a critical feature for marketing and creative production. If you generate something you like, being able to reproduce that image in different settings or even poses is something that’s been lacking. I loaded the image I created in Photoshop into the new reference image panel:
Here are the results:
I think it did a great job recreating the style and tone of the image, but it did not accurately re-create the character in a different pose. I’ve seen this work much better in Dalle-3 and other image creation tools such as Midjourney. It does say that its goal is to match the style, but I can always hope. 🙂
