Canva Launches AI Image Generator

Canva Launches AI Image Generator

by

Australian design platform Canva released its first in-house image generation model, Canva Craft, developed to compete with major AI labs, the model uses a two-stage architecture to plan layouts before rendering images up to 2K resolution, arriving amid cost-cutting pressures and valuation shifts.

Canva has entered the foundational AI model race with the release of an in-house image generator designed to address persistent flaws in digital design workflows. The software giant launched Canva Craft following an emergency reworking of its artificial intelligence tools. The company had previously been forced to halt the launch of high-profile AI features in August due to the unsustainable token costs of running third-party models for non-paying users, a delay that coincided with backers slashing $10 billion from valuations that now sit at $50 billion, according to financial reporting and external analysts.

How Canva Craft Plans Compositions Before Rendering

Built from scratch by Canva Research—an in-house lab comprising more than 160 scientists and engineers—Craft relies on a distinctive two-stage architecture. In the initial phase, the model interprets user prompts to draft a structural plan accounting for spatial relationships, relative scale, visual hierarchy, and text generation requirements. Only after this compositional planning does the second stage render the final output at resolutions up to 2K across more than 30 aspect ratios.

Rather than drawing primarily from photographic data, Canva trained the model on millions of human-designed templates from its own library, including posters, social media graphics, presentations, infographics, menus, and typography-rich layouts. According to the company, this design-first training approach allows the generator to produce accurate object placement and legible text while minimizing the awkward layouts common to general-purpose image tools. Users can also execute targeted edits without altering unrelated sections of the image.

Canva Launches AI Image Generator
Photo: Traders Union

Canva Craft Ranks Second Globally in AI Image Generation

Initial benchmark results placed Canva Craft fourth on Design Arena, a crowdsourced evaluation platform where images generated from identical prompts undergo blind comparisons. That performance positions Canva as the second-highest-ranked AI laboratory for image generation globally, trailing only OpenAI.

Independent tests cited by regional reporting indicate that Craft outranks competing models from Google, Microsoft, Meta, xAI, and ByteDance. Specific category breakdowns on Design Arena place the model first for generating images of people and natural landscapes, while securing second, fourth, and fifth positions for abstracts, products, and marketing materials respectively, alongside a 14th-place finish out of 87 for cartoons.

Integration with Dream Lab, APIs, and Workflow Tools

The new model is currently accessible to all users globally through Canva’s left-hand menu under Dream Lab, as well as via the company’s public API for external developers. Canva also intends to support image generation within third-party chatbots such as ChatGPT and Gemini through the open Model Context Protocol. Furthermore, the technology integrates with Canva Magic Layers, an in-house feature introduced last year that converts generated images into multi-layered designs rather than leaving them as flat files.

Behind the launch lies an economic imperative for the Australian software unicorn. Developing proprietary foundational models provides direct control over generation speed, quality, and operating costs, reducing reliance on expensive external APIs. The company’s research foundation also draws on talent from Leonardo.ai, a media generation platform acquired by Canva in 2024. Additional developments are already underway, with the company preparing to release two more cost-efficient versions named Craft-fast and Craft-ultrafast.

Canva craft official launch artwork
Photo: Kingy AI

Strategic Growth Objectives and Cost Control

With artificial intelligence capabilities becoming increasingly central to its overall product strategy, Canva’s immediate objective is to make AI-generated visuals easier to produce, edit, and incorporate into finished designs while taking greater charge of technology expenses. External analysts have noted that despite earlier tracking toward $4 billion in revenue, the company adjusted its growth forecast for 2026, which contributed to valuation adjustments from external market observers. By deploying Canva Craft, the business secures a proprietary pathway to serve the hundreds of millions of monthly users without prohibitive inference fees.

In an official statement regarding the capabilities of the newly introduced architecture, JJ Fiasson, head of generative AI at Canva, remarked that the technology delivers exceptional results for tasks requiring precise text and layouts while rendering landscapes and people with stunning quality. Fiasson added that additional steering capabilities and major speed improvements are on the way, representing a significant step forward for the platform’s overall artificial intelligence experience.

>

Worth a look

Read next