Adobe Demos New AI Generating Tool That Turns 3D Shapes Into Desired Images
Video screenshots via Figma
Adobe is taking the abilities of its Firefly generative artificial intelligence to the next dimension, as previewed at a recent conference in San Francisco.
Following the much raved-about rollouts of Generative Fill for Photoshop and Generative Recolor for Illustrator, Adobe revealed at the 2023 Figma Config last week that it’s working on another art-generating AI tool, codenamed ‘Gingerbread’, that uses 3D models to guide the machine to deliver precise and desirable imagery.
As it turns out, you can replicate many objects in this world from the same set of shapes. For instance, at the demonstration, Adobe chief product officer Scott Belsky and principal designer for emerging design Brooke Hopper used a 3D cylinder to dream up a render of a whisky glass, assisted by a text prompt.
Prompt: “Whisky glass on a wooden table”
Video screenshots via Figma
Video screenshots via Figma
Gingerbread can also determine the style, depth, and perspective of a setting, and adjust the object to make the entire scene more photorealistic.
The feature is envisioned to come under the Firefly wing, the company’s AI-powered pride and joy that’s trained on public-domain, Adobe Stock, and openly-licensed images.
Prompt: “Gold headphones, inside empty warehouse, clouds of colored smoke float in air, bokeh”
Video screenshots via Figma
The thing with many existing AI generators is that, regardless of the prompt you use, you can’t fully predict what the bot will churn out. Gingerbread aims to give creatives more control in this area as allowing them to set down placeholders, resize them, change their perspectives, and more.
You can watch the full preview below.
It remains to be seen if Adobe will eventually bake Gingerbread into reality. Still, the experiment hints at how expansive the world of generative AI can get, and that we’re only just scratching the surface.
this is amazing.@adobe just showcased project ''gingerbread'' at #Config2023 — a image generative AI using 3D objects to tailor your image.
— phil desforges (@storybyphil) June 22, 2023
Here, Brooke Hopper used a simple cylinder and a text prompt ''whisky glass'' to create this. pic.twitter.com/a1BI3fZ1A5
