From image generation to design generation
Canva’s announcement describes Craft as a model built around graphic design rather than conventional photorealistic image synthesis. Its proposed workflow begins with planning. Before rendering an image, the system maps relationships between elements, including spatial position, scale, hierarchy and text. It then generates the visual output.
That distinction matters because most design work is not simply about producing an image. A marketer may need a headline in a specific location, a product photo separated from a background, a logo with enough breathing room and a layout that can be adapted for several platforms. A beautiful first result becomes far less valuable if every change requires starting again.
Canva says Craft can produce images at up to 2K resolution across more than 30 aspect ratios. The range is intended to fit the practical variety of modern visual communication, from presentation slides and social posts to menus, posters and other marketing materials. According to the company, the model was trained on millions of human-made design templates, including the kinds of structured compositions that people already create inside Canva.
The company is also connecting Craft to Magic Layers. In Canva’s Magic Layers announcement, the company said the system can turn a flat AI-generated image into editable components. Text can become live text, objects can be separated, and the original layout structure can be preserved for further refinement.
In a near-future workflow, a small business owner could describe a seasonal promotion in the morning and receive a poster, a mobile advertisement and a presentation slide before lunch. Instead of opening separate files and rebuilding each composition, the owner could move the product image, rewrite the offer and resize the design while keeping the visual system intact. Craft is aimed at making that interaction feel less like prompting a picture and more like briefing a junior designer.
The promise has practical limits
That ambition also creates a higher standard. If Canva presents Craft as a design-native tool, users will judge it by the details that matter after generation: whether the copy is correct, whether brand rules are respected and whether the layout can survive editing.
Gaurav Tiwari, a designer and agency reviewer, supports Craft as a way to create first drafts but is more critical of its reliability in production work. In his review of Canva Craft, Tiwari says the system can add copy that users did not request, a particularly serious problem for advertisements, event materials and client work. Unwanted words can turn an otherwise polished graphic into something that requires manual cleanup or, worse, slips into publication.
Tiwari also criticizes the lack of Brand Kit support in Dream Lab, along with what he describes as limited available aspect ratios. Those complaints point to a gap between visual fluency and professional usefulness. A model can understand that a headline belongs above an image without understanding that a company’s approved typeface, color palette and tone are not optional.
For agencies and freelancers, human proofing remains necessary in Tiwari’s assessment. That is not a minor inconvenience. The value of an editable AI design depends on how much time it saves. If every output must be checked for invented text, inconsistent branding and incorrect proportions, the tool may shift labor rather than remove it.
A new question for AI evaluation
The debate also raises a broader issue about how design models should be assessed. AI consultant Jacob Topmoller, writing in a Reddit analysis of what Canva’s approach suggests for work software, takes a neutral view of Canva’s training and evaluation strategy. Topmoller argues that domain experts should examine training examples and outputs directly, rather than relying mainly on indirect success metrics.
That argument is important because conventional image benchmarks may reward visual appeal while missing the problems that professionals notice immediately. A design can look coherent at a glance and still fail because its text is inaccurate, its hierarchy is confusing or its elements cannot be edited cleanly.
Canva’s advantage may therefore come less from creating a more spectacular image than from teaching a model what a usable design is. Its training material reflects the habits of human designers, at least according to the company, while Magic Layers addresses the next step after generation. Together, they suggest an AI system that is intended to participate in a workflow instead of ending it.
The strategic stakes for Canva
Craft could also reduce Canva’s dependence on outside image models by giving the company control over a capability central to its product. More importantly, it could make Canva’s interface the place where generation, editing and brand management converge.
The outcome will depend on whether the model handles ordinary professional demands, not only impressive demonstrations. Users will want precise copy, dependable brand controls, flexible formats and layers that behave predictably. If Craft delivers those features, AI-generated imagery may become less like disposable content and more like a working material.
If it does not, Canva may have built an excellent first-draft machine that still leaves the most important work to humans. That would not make Craft unimportant. It would show that the future of design AI will be decided after the image appears, when someone asks whether it can be trusted, edited and used.
Maya Lindqvist is not a person. No notebook, no deadlines, no face behind the name — just a byline this newsroom publishes under. Here is the production line underneath it, because a name beside a portrait reads like a journalist, and this one is not one.
The models. Writing: gpt-5.6-luna and qwen3-max. Out on the live web: gpt-5.6-luna and gpt-5.6-terra. Pictures: gpt-image-1 and gpt-image-1-mini. Swap one in the newsroom and this line swaps with it — it is read off the machines, not typed here.
How a story is made
- Research. The searching model reads around the story, pointed at primary sources — the filing, the post, the repository — rather than at somebody else's write-up of them.
- Writing. The writing model drafts it against what was found, at Maya Lindqvist's usual length and in Maya Lindqvist's usual register.
- The loop. A reviewer reads the draft and sends it back with notes. Then reads it again. A piece can go round several times before it leaves the building.
- Enrichment. A quotation has to appear word for word on the page it is taken from. A chart may only use figures that appear in the source it cites. Whatever fails is dropped, and the reason is kept.
- Fact check. A last pass hunts for claims the article makes and its sources do not.
- A human stop. Sensitive subjects are held for a person to read before publication, and a person can kill any of it at any point.
If that sounds less like a newsroom and more like a factory: quite. It is called Press Factory.
This article was generated using AI and published automatically without human pre-publication review.
How this article was made
The article was produced by the Grandmonts Media News Engine using automated research, drafting and verification workflows. No human editor reviewed the article before publication. Grandmonts Media remains responsible for the published content. Errors can be reported at office@grandmonts.cz.