Variation is about to be free. The brief, and the eye that catches a wrong edit, are not.
Canva Brush for designers: the short answer
Canva released Canva Brush on 29 October 2026, and the practical summary of Canva Brush for designers is this: it is an editing model, not an image generator. According to Canva's official Brush announcement, Brush is a vision-language model built by Canva Research that "sees your design, understands how it's put together, and makes the changes you ask for", from one line of copy to every slide in a deck. It runs inside Canva AI for Canva Pro subscribers, with wider availability described as coming soon.
The distinction from last week's model matters. Canva Craft generates an image; Brush edits the design around it. If you already read the breakdown of Canva Craft for designers, think of Brush as the second half of that story: the part where a generated composition becomes something you can revise by prompt without losing the elements you did not mention.
What Canva actually shipped
Kept to what Canva states in its announcement:
A model for edits, not renders. Brush understands the structure of an existing design (text, imagery, colours, layout) and applies a requested change while keeping the rest intact. Canva frames this as the harder problem: an edit has to know what to change and, just as importantly, what must stay exactly as it is.
An agentic loop. Behind the scenes, Brush writes and runs code to adjust design elements, then inspects what happened on the page to guide its next step. That is why it can take a multi-step request, such as a three-page deck with a consistent visual style, and keep typography, colour and spacing coherent from the first page to the last.
Fully editable output. Everything Brush produces stays a normal Canva design. Every element remains yours to move, restyle or delete by hand.
Efficiency claims. Canva states that Brush is up to 69 percent faster on general design editing tasks and 46 percent lower in cost per edit task than its previous design agent. Those are Canva's numbers from Canva's own comparison, so treat them as a direction, not an independent benchmark. Canva also says its internal testing showed Brush matched or exceeded the previous setup on most editing tasks, with the strongest results in restyling designs and adapting them to other formats.
How it was built. Brush started from a leading open-weights model, was post-trained on real design and editing tasks, then shaped with large-scale reinforcement learning against a reward system that scores visual quality, adherence to layout principles and whether the task was actually completed.
Access. Canva Pro plans from launch day, inside Canva AI. Expanded availability is promised without a date.
What is not there yet. Canva lists Brand Kit support, referencing attachments, and better handling of image-heavy layouts as work in progress. Read that list as the honest map of where the first version is weakest.
How to use Canva Brush: a working loop
Canva's own instructions are short: open an existing design in the editor, or attach or link a design in Canva AI, then select Brush from the model dropdown in Canva AI so it powers the full editing session. From there you alternate between prompting and editing by hand.
That is enough to start, but not enough to ship. A better routine for client work is a four-step loop: Lock, Ask, Inspect, Finish.
Lock. Before prompting, decide what must not change: the logo, the legal line, the brand colours, the approved headline. Write those into the prompt explicitly ("keep the logo, footer and headline exactly as they are"). Brush is trained to preserve what you do not mention, but naming the untouchables costs nothing and protects the brief.
Ask. Give one edit per prompt, scoped to a page or an element set. Canva's three launch examples are the right size: restyle an image or a whole design in a different direction (a new colour scheme, different weather, a different mood), change something across every page of a multi-page design in one request, or remix an asset for another channel, such as turning a presentation slide into an Instagram graphic.
Inspect. Check the result the way a reviewer would, not the way a prompt author would. Did spacing drift? Did text reflow and break a line in the wrong place? Did a restyle touch a colour that was supposed to be locked? Because the output stays editable, you can select any element and see exactly what Brush did to it.
Finish. Take over in the editor for the last five percent. Kerning, final crop, alignment to the grid, the export. Canva's announcement treats the hand-off to manual editing as part of the workflow, not a fallback.
Prompts that tend to work with editing models are specific about scope and outcome and silent about method. "Change every page header to the deep green from the palette and leave body text as it is" beats "make it look more premium". If a result misses, undo and rewrite the scope rather than piling a corrective prompt onto a wrong state.
Canva Brush vs Canva Craft vs editing by hand
Criterion | Canva Brush | Canva Craft | Manual editing in Canva |
|---|---|---|---|
Job | Edit and restyle an existing design | Generate a new image or composition from a prompt | Precise control of every element |
Input | A design plus a prompt | A prompt (reference images announced as coming) | Your hands and the editor |
Output | The same design, changed, fully editable | A new image, convertible to layers with Magic Layers | Whatever you build |
Best at (vendor stated) | Restyling, multi-page changes, reformatting for channels | Planned layouts with legible text, 30+ ratios, up to 2K | Final polish, brand-critical details |
Where it runs | Canva AI, Canva Pro at launch | Dream Lab for all users, public API | Any Canva plan |
Brand Kit support | Announced as coming | Not stated | Full |
Risk to watch | Unrequested changes on image-heavy layouts | Fine print and technical drawings | Time |
Pick Brush when a design exists and the job is variation, adaptation or a sweeping change. Pick Craft when nothing exists yet and you need a starting composition. Keep manual editing for anything a client will inspect at 200 percent zoom.
Hype vs reality
The hype: Canva now has its own frontier design lab, two in-house models in two weeks, and an editor that reads your layout and fixes it on command. Part of that story leans on a DesignArena ranking Canva cites, which is a crowdsourced preference vote, not a test of whether a deck survives a brand review.
What holds up: editing is the right problem to solve. Most AI design tools fail not at generation but at the second request, when a small change ripples through things you did not ask to touch. A model trained specifically to preserve structure, and that inspects its own output before replying, is aimed at that failure. Keeping everything editable is also the right call. A locked raster is worthless in production.
What does not hold up yet: no independent evaluation, no Brand Kit support at launch, no attachment references, and Canva's own admission that image-heavy layouts need work. For a designer whose output is mostly photo-led campaign assets, that last point is the one that matters.
Verdict: worth a serious trial this week for teams that produce multi-page, multi-format work in Canva (decks, social sets, menus, event collateral). Not yet a substitute for a brand system, and not something to point at a client-facing file without the Inspect step.
Where this takes design work
Three things change over the next couple of years as editing models like Brush get better and cheaper.
First, variation becomes free, so the value moves to the decision. Twenty format adaptations of a campaign used to be a day of production. When they take minutes, the paid skill is choosing the right twenty and knowing which three are wrong. That favours designers who can articulate rules, which is why writing design systems as AI context is becoming a core craft rather than a documentation chore.
Second, brand rules will be enforced by models before they are enforced by people. Canva's roadmap puts Brand Kit support next, and tools like the ElevenLabs Brand Kit are moving the same way. The designer's job shifts toward defining the kit precisely enough that an agent cannot misread it.
Third, the editor becomes a place you supervise as much as a place you draw. Prompt, inspect, correct, hand-finish. That loop looks a lot like reviewing a junior's work, and the designers who are already good at giving clear, scoped feedback will be the ones who get the most out of it.
The fear, answered
Will Canva Brush replace designers? Not the ones whose value is judgment. It replaces the hours spent propagating one decision across forty artboards, and it lowers the floor for people with no training, which means more mediocre design in the world, not less demand for good design. The honest risk is for roles that were mostly production: resizing, retyping, restyling to spec. If that is most of your week, the move is to own the spec. Learn to write the rules, build the templates, run the review. The longer argument is in will AI replace UX designers.
Am I falling behind if I do not use it? No. You are falling behind if you cannot explain to a client why an AI edit is wrong and what the fix is. Try Brush on a real multi-page file this week, note where it drifts, and keep that list. That is the knowledge clients pay for.
FAQ
What is Canva Brush and who can use it?
Canva Brush is a vision-language model for design editing, built by Canva Research and released on 29 October 2026. It powers supported design edits in Canva AI for Canva Pro subscribers, with expanded availability announced as coming soon. You select it from the model dropdown in Canva AI and it edits an existing design while keeping every element editable.
How is Canva Brush different from Canva Craft?
Craft generates a new image or composition from a prompt. Brush edits a design that already exists: restyling it, changing something across every page, or adapting it to a new format. Canva's own summary is that Craft generates the image and Brush refines the design around it.
Is Canva Brush overhyped?
The underlying idea is sound: an editing model that preserves what you did not ask to change is what designers actually need. The speed and cost figures are Canva's own comparison with its previous agent, and there is no independent evaluation yet. Brand Kit support and attachment references are not in the first version. Treat it as a strong production tool to trial, not a finished brand guardian.
Will Canva Brush replace graphic designers?
It replaces repetitive production work, not design judgment. Deciding what a brand should look like, briefing the edit correctly, catching a wrong change and finishing by hand remain human work, and they are where the value sits as variations get cheap.
What should I check before shipping a Brush edit?
Locked elements (logo, legal copy, brand colours), text reflow and line breaks, spacing against the grid, colour values against the palette, and image crops on photo-led pages, which Canva says it is still improving. Export, then review the export, not the canvas.
Where to get help
Tools like Brush make the edit cheap and the brief expensive. If your team needs the brief done properly, from brand rules an agent can follow to the screens and assets behind a launch, DIGCY is a product design agency in Casablanca working with clients globally, offering UI/UX design services and motion design and AI video for teams shipping on AI-assisted timelines. The studio is a Dribbble Selected Agency with 50+ projects delivered. Start with one real file: run it through the Lock, Ask, Inspect, Finish loop and keep notes on where it drifts.
Let’s keep in touch.
Discover more about high-performance web design. Follow us on Twitter and Instagram.



