canva brush Model: 69% Faster Design Edits for Pro

The canva brush model is the design platform's first vision-language model, and it went live on October 29 for Canva Pro subscribers inside Canva AI. Built by Canva Research, it reads an existing design, figures out how it is put together, and edits it on request, from one line of copy to every slide in a deck. Canva says it is up to 69% faster on general editing tasks and 46% cheaper per edit than its previous design agent.

What actually happened

Canva announced Brush on its newsroom as "a new vision-language model for design." In plain terms, a VLM can both look at an image and follow written instructions about it. Brush is tuned for the structure under a design: layers, layout, typography, and style.

The pitch is editing, not generation. Craft generates images; Brush refines the design around them. Canva frames the hard part as knowing what must stay the same. Ask a generic model to change one header and it may also wreck your spacing, colors, and fonts. Brush is meant to make the requested change and leave everything else alone.

Under the hood, Canva says Brush runs an agentic loop. It writes and executes code to adjust elements, checks what happened on the canvas, and then decides what to do next. Everything it touches stays fully editable, so you can take over by hand at any point.

How Canva built it

Canva says canva brush started from a leading open-weights model, which the team post-trained on real design and editing tasks. Training also drew on anonymised insights from people who opted in to help improve Canva AI. On top of that, Canva ran large-scale reinforcement learning with a reward system that scores visual quality, layout against design principles, and whether the task was completed as asked.

The company says professional designers on its team set the definition of "good." In internal testing, Brush matched or beat the previous setup on most editing tasks, with the strongest results in restyling designs and adapting them to new formats. Those are Canva's own numbers, and no outside benchmark has been published for Brush yet.

Craft makes the picture. Brush fixes the deck around it.

Capital Brief and Crypto Briefing both flagged the same strategic shift: Canva is moving from renting general-purpose AI to building its own specialised models, with a 160-person research lab behind both releases. Crypto Briefing also noted that training on anonymised user data could draw scrutiny, a recurring debate across the industry.

Why the canva brush launch matters

Efficiency is the real headline. Canva runs at the scale of hundreds of millions of users, and a few extra cents per AI task adds up quickly. A 46% cut in cost per edit gives Canva room to ship more AI features without eating margin at the same rate. For users, the 69% speed claim means less waiting between prompts.

It also fits a wider pattern we have tracked. Our canva craft piece covered the image model that came first, and Microsoft's mai image 2.7 preview shows big platforms racing on editing quality, not just generation. The canva brush release pushes that race into layouts, decks, and multi-page designs.

What to do as a reader

If you are on Canva Pro, open an existing design, go to Canva AI, and pick "Brush" from the dropdown. Canva suggests three starter jobs: restyling an image or design, changing headers across a multi-page deck in one prompt, and remixing a slide into a new format such as an Instagram post.

Test canva brush on a copy of an important file first. Check that fonts, brand colors, and spacing survive the edit. Brush is a first version, and Canva says Brand Kit support and better handling of image-heavy layouts are still coming.

If you build creator tools or design-adjacent products, watch pricing and availability. A cheaper in-house editing model at Canva scale can reset what users expect from AI design features everywhere.

Sources: Canva newsroom, Capital Brief, Crypto Briefing.