Canva Craft Explained: The Design Platform’s First In-House AI Image Model

Canva has released Craft, its first in-house image generation model, built for design work rather than photographs. It plans first, then renders, and its output can be turned into editable layered designs. Canva claims it is now the world’s second most advanced AI lab for image generation, which is the company’s own framing.

What Makes Canva Craft Different

Most image models turn a prompt directly into pixels. Craft uses two stages: it first plans the composition, including positions, sizes, visual hierarchy and text, and then renders the image. Canva says it trained on human-made designs from its template library, not mainly on photos.

Details are from Canva, ProPakistani and Thurrott.

How Good Is It?

Canva says Craft ranked fourth among image models on DesignArena, a crowdsourced benchmark, behind three OpenAI models. Because the leading lab places several models above it, Canva’s “second most advanced lab” line counts labs, not models. It is a company claim and the leaderboard can shift quickly.

Why Canva Built It

• Cost: Canva has relied on several outside image models, so its own could cut API costs, though it hasn’t disclosed savings.
• Control: It can tune quality and speed for design use.

Read about Canva. More AI news is on TechnoSports.

Frequently Asked Questions