Instead, they generate flat, JPEG-style outputs in which text, backgrounds, and other elements cannot be separated. With Canva’s design model, it’s possible to generate “layered design files that look amazing, but then then allow for AI and human editing to happen in parallel,” said Thompson.
Once a design is generated, it can be tweaked as with any template in the Canva editor.
The model enables an “array of functionality, both today and especially in the future,” said Gabe Knuth, principal analyst at Omdia, part of Informa TechTarget. “Using their own model will allow [Canva] to have more control over the functionality and output than they’d have leaning on different foundational models for different tasks.”
This story originally appeared on Computerworld
