Canva launches a home-grown AI image generation model
Article excerpt
If you like this article, share it with your friends. AI/Machine Learning Business Share Canva says it plans composition, layout and text before generating an image, aiming to produce legible lettering and designs that need fewer corrections. The company said that through advanced model distillation, it reduced the number of inference steps required by Craft to a fraction of the original model. The launch comes 10 weeks after Canva CEO and cofounder Mel Perkins flagged a cut in Canva’s forecast revenue growth by a third, from 30% to 20%, as the business tackled escalating AI costs. Perkins said in early August that the “average cost of serving an AI task was too high” and the design platform was “relying too heavily on frontier models”. Two weeks later, Canva’s valuation was cut by $11 billion, a drop of 20%, to US$31bn (A$43.5bn), down from US$38.9bn a year ago. The company was also dealing with Now the company’s entering a crowded battlefield hoping that its home-grown AI is a better, cheaper alternative to the frontier models by tech giants such as OpenAI’s ChatGPT, Google’s Nano Banana and Adobe Firefly. Building its own model gives the business greater control over costs, although development and computing remain serious expenses. Canva is also developing faster, cheaper Craft variants and plans connections that let AI assistants use the model. By continuing, you agree...
Keep reading with a free account
The rest of this article, and every signal for Canva, is in your free account.
Extracted from this sentence
Fraudsters exploited linkages between Canva and Leonardo.ai to acquire free tokens and resell them.
