Gemini 3.5, GPT 2.5 and Reverse layering
Gemini 3.5, GPT 2.5 and Reverse layering
September 08, 2026

Today OpenAI announced GPT 2.5 Image, and we’ve rolled it into PictureStudio. Like Google’s Gemini 3.5, it’s a highly capable frontier model that works two ways: it can generate images from a description, and it can edit images you give it. Both models are capable enough that they open up new ways of working.
One technique we’ve seen artists using is something we call reverse layering. You start with a complete image, like the one above. Then you write a prompt to isolate just the character, placing it against a green screen.

From there you can generate character sheets.

Change outfits.

Or build sprite sheets for animations. The trick with sprite sheets is to describe the purpose of the animation, then each frame in detail (this is a 6 frame walking animation that loops, repeating perfectly. Frame 1:...).
With workflows like these, you can move from a rough idea to production-ready assets at a speed that didn’t seem possible even a few months ago.
If you try these techniques, tag us on social media. We'd love to see what you create!