{"slug": "canva-unveils-brush-its-first-vision-language-model-built-for-design", "title": "Canva unveils Brush, its first vision-language model built for design", "summary": "Canva launched Canva Brush on October 29, 2026, its first in-house vision-language model, built to edit designs alongside users and available at launch to Canva Pro subscribers. Canva says Brush makes design editing up to 69% faster and costs 46% less per editing task than its prior agents, with outputs that stay fully editable. The model comes from Canva Research, Canva's 160-person AI lab that also built the Canva Craft image generation model, and was trained on real design tasks and anonymized user data.", "body_md": "# Canva unveils Brush, its first vision-language model built for design\n\nThe in-house model reads layouts, layers and styles to edit designs alongside users, and it debuts for Canva Pro subscribers\n\nCanva has a new collaborator, and it doesn’t need a coffee break.\n\nThe design platform launched **Canva Brush** on October 29, 2026. It is a vision-language model built to edit designs alongside the people making them.\n\nThat matters because Brush is homegrown. Canva is moving away from renting AI and toward building the engine itself.\n\n## What Brush actually does\n\nA vision-language model, or VLM, is AI that can both see and read. Think of it as a model that looks at a slide the way a designer would, then follows written instructions about what to change.\n\nBrush is tuned specifically for design work. Canva says it understands layouts, layers and visual styles, the structural bones that sit under any poster, deck or social graphic.\n\nUsers can ask for small or sweeping changes. The range runs from tweaking a single line of text to redesigning an entire presentation.\n\nThe key detail is what happens after the edit. Brush’s outputs stay **fully editable**, so text, imagery and layout remain under the user’s control.\n\nThe model also runs a feedback loop. It analyzes the live design canvas as it works and refines its output based on what it sees.\n\n## The numbers behind the launch\n\nCanva is pitching Brush on speed and cost. The company says design editing runs up to 69% faster compared to its prior agents.\n\n### AI, tech, and the markets they move—in one daily briefing.\n\nDaily. Free. Join 34,000+ readers across crypto, finance, and policy.\n\nEditing tasks also cost 46% less relative to previous models.\n\nBrush comes from Canva Research, the company’s 160-person AI lab. The same team built Canva Craft, an image generation model released shortly before Brush.\n\nCraft makes images. Brush rearranges and refines the designs those images live in.\n\nTraining drew on real design tasks and anonymized user data. The goal, according to Canva, is to sharpen the editing experience for everyone on the platform.\n\nAt launch, Brush is available to **Canva Pro** subscribers.\n\n## Why Canva is building its own models\n\nCanva is taking a different path. Brush and Craft both reflect a shift from relying on outside models to creating proprietary systems designed around design workflows.\n\nThe company is also building for interoperability with external AI platforms.\n\n## What this means for Canva and its rivals\n\nThe most direct effect is likely on Canva Pro. Gating Brush behind the paid tier gives subscribers a concrete reason to upgrade, especially professionals and teams who edit large volumes of material.\n\nIf editing tasks cost 46% less to run, Canva may be able to offer more AI features without seeing margins erode at the same pace.\n\nThe use of anonymized user data in training could also draw scrutiny. That kind of approach has become a recurring point of debate across the AI industry.\n\nStarting with Pro users gives Canva a controlled audience for feedback.\n\nFor now, Brush marks a clear statement of intent. Canva wants its AI to understand design the way its users do, layer by layer, and it has decided the best way to get there is to build it in-house.\n\n**Disclosure:** This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our [Editorial Policy](https://cryptobriefing.com/editorial-policy/).\n\n## More in Technology\n\n[See all](https://cryptobriefing.com/tech/)", "url": "https://wpnews.pro/news/canva-unveils-brush-its-first-vision-language-model-built-for-design", "canonical_source": "https://cryptobriefing.com/canva-unveils-brush-vision-language-model/", "published_at": "2026-10-29 19:50:14+00:00", "updated_at": "2026-10-29 20:20:21.229310+00:00", "lang": "en", "topics": ["artificial-intelligence", "generative-ai", "ai-products", "computer-vision", "large-language-models"], "entities": ["Canva", "Canva Brush", "Canva Research", "Canva Craft", "Canva Pro"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/canva-unveils-brush-its-first-vision-language-model-built-for-design", "markdown": "https://wpnews.pro/news/canva-unveils-brush-its-first-vision-language-model-built-for-design.md", "text": "https://wpnews.pro/news/canva-unveils-brush-its-first-vision-language-model-built-for-design.txt", "jsonld": "https://wpnews.pro/news/canva-unveils-brush-its-first-vision-language-model-built-for-design.jsonld"}}