# What is Canva Brush? Canva's vision-language model for design

> Canva Brush is a vision-language model from Canva Research that understands how a design is built and edits it on request, from one line of copy to a…

Source: https://sketchry.app/blog/canva-brush/

AI tools · 2026-10-29 · Source [canva.com](https://www.canva.com/newsroom/news/canva-brush/)

Canva Brush is a vision-language model from Canva Research that understands how a design is built and edits it on request, from one line of copy to a whole deck.

Canva Brush is a vision-language model Canva announced on October 29, 2026, built by Canva Research, the company's 160-person AI lab. Brush looks at your design, understands how it is put together, and makes the changes you ask for, from a single line of copy to every slide in a presentation.

Canva says Brush is up to 69% faster on general design editing tasks and has 46% lower cost per edit task than its previous design agent. It works through an agentic loop that writes and runs code to adjust design elements, then inspects the page to decide its next step.

Everything Brush creates stays fully editable, so text, imagery, colours and layout can be changed by hand like any other Canva design. Suggested uses include restyling a design or image with a new colour scheme, weather or mood, updating headers across a multi-page design in one prompt, and resizing a slide into an Instagram graphic.

Brush started from a leading open-weights model that Canva post-trained on real design and editing tasks, followed by large-scale reinforcement learning. Its reward system scores visual quality, how well the layout follows design principles, and whether the task was completed, with the standard for good design set by professional designers on the team.

Brush now powers supported design edits in Canva AI for Canva Pro subscribers, with wider availability coming later. To use it, open a design or attach one in Canva AI and pick Brush from the dropdown. Canva says Brand Kit support, attachment references and better handling of image-heavy layouts are in the works.

Last updated: 2026-10-31
