Canva Craft AI Image Model: What It Means for Cutouts [2026]

Canva Craft, the company's first in-house AI image model, launched on October 22, 2026, and it is available to all Canva users through Dream Lab. The headline feature is not raw image quality. It is that Craft plans a design before it draws one, then lets you split the result into editable layers. For anyone who cuts out subjects, swaps backgrounds or builds social graphics, that is the part worth understanding.

This post sticks to what Canva, The Verge, TechCrunch and the Product Hunt launch page actually report. We have not benchmarked Craft ourselves, and nothing here is a hands-on test result. Canva's own newsroom page would not load for our fetch, so every Canva claim below is as relayed by those outlets.

The short version for creators: Craft is a generator, not a background remover. It changes how you build a composition from scratch, and it makes layered, text-heavy designs easier. It does not, according to any coverage we found, process video. If your work is cutting people out of footage or cleaning product photos, the workflow questions are different, and we cover them below.

What is Canva Craft and when did it launch?

Canva Craft is Canva's first AI image generation model, developed by its own research team and announced on Thursday, October 22, 2026. TechCrunch frames it as a model built to compete with same-category models from Google, OpenAI, ByteDance and Meta, and The Verge says it competes directly with those from OpenAI, Microsoft and Google.

The model produces images at up to 2K resolution in more than 30 aspect ratios, per The Verge. That aspect-ratio range matters for creators because it covers stories, banners and posts without cropping afterwards. Product Hunt's launch description adds that it renders legible text in 55 languages and that outputs can be split into editable layers. Canva also lists an API, so developers can call it outside the editor.

Canva says it trained Craft on "millions of human-made templates" from its library: posters, social posts, presentations, infographics, menus and typography-heavy layouts. That is a different diet from models trained mostly on photographs. TechCrunch also notes that researchers from Leonardo.ai, which Canva acquired in 2024, contributed to the model.

Where do you find it?

According to The Verge, Craft is available now to all users globally. Open the three-dot option on Canva's left-hand menu and choose Dream Lab. Canva has not, in the coverage we reviewed, published usage limits or credit costs for Craft, so check your own plan before assuming it is unlimited.

How does Canva's two-stage design actually work?

Craft first writes a plan, then renders from it. TechCrunch describes the first stage as interpreting the prompt and creating a plan with spatial relationships of elements, relative scale, visual hierarchy and information for text generation. The second stage generates the image, up to 2K, from that plan.

Canva's own words, as quoted by The Verge, are that this allows "greater precision, consistency and control." In practice the claim is about two failure modes every AI image user knows. The first is text that comes out as alien letters. The second is objects that land somewhere other than where you asked. A model that decides where the headline and the product go before it paints pixels should, in principle, do better on both. The Product Hunt hunter who introduced the launch makes the same point from experience: their pet peeve with AI images is text, and they report that in Craft the text reads like text and things go where they were asked to go. That is one user's impression on a launch page, not a measurement.

What does planning change for backgrounds?

Planning makes the background a deliberate element rather than a by-product. If the plan separates a subject, a backdrop and a text block, then asking for "the same product on a different background" is closer to editing one element than rerolling the whole image. TechCrunch reports that Canva says users can make precise edits without changing other parts of the image. We have not tested how reliably that holds up, and a generated background is still a generated background: it does not match the lighting of a real photograph you shot yourself.

Can you turn Craft images into layers and cutouts?

Yes, with Canva's existing Magic Layers tool, according to TechCrunch. Because Craft learned from designs rather than photographs, TechCrunch says its images "could be converted into layered designs using Magic Layers, the company's image-to-design transformer."

The Verge adds context: in April, Canva launched prompt-based editing and brought layers to AI-generated images, letting you change part of an image while leaving the rest untouched. Craft slots into that workflow as the generator at the front.

What layers do and do not give you

Layers let you move, resize or delete one element without regenerating the picture. That covers many tasks people do with background removal on posters and social graphics: swap a backdrop, reposition a product, change headline text. What layers do not obviously give you is a clean matte of a real person with wispy hair against a busy photo. That job depends on edge quality, which is a segmentation and matting problem. Our hair and edge guide explains why those edges are hard, and the same caution applies to anything that splits a flattened image into pieces.

If your end goal is a transparent file, our walkthrough on how to make a transparent background online covers the export checks, and the image background remover handles single-photo cutouts directly.

Where does Canva Craft rank against GPT-Image?

Fourth on Design Arena, a crowdsourced benchmark for AI design images, according to The Verge. It sits behind OpenAI's GPT-Image-2.5 Sunburst, GPT-Image-2.5 Flare and GPT-Image-2. That is a leaderboard of crowd preferences at launch, not a test of cutout quality, and rankings move.

Treat that last row carefully. Absence from coverage is not proof Canva cannot do it, but nothing we read says Craft generates or edits video.

Why is Canva building its own model?

TechCrunch's read is that Canva has used many third-party image models and that its own model should save API costs, plus give it better quality and speed control. It also notes Canva cut its 2026 growth forecast, which analysts reportedly linked to a valuation drop from 42 billion to 35 billion dollars. That is TechCrunch's reporting, and it frames the launch as part of a margin story as much as a creative one. For users, the practical effect is that Canva can tune one model to its design use case instead of renting a general one.

Does Canva Craft matter if you work with video?

Mostly not yet, and that is the gap to be clear about. Everything announced concerns still images: generation, layers, text rendering. Neither The Verge nor TechCrunch describes motion or video output.

A moving subject needs a different kind of consistency. Each frame's mask has to agree with the previous one, edges must not flicker, and audio must stay aligned. We walked through the model side of that in our SAM 3 video background removal guide, and the real-time angle in interaction models and the future of video background removal.

If you generate a clip with an AI video model and then want it on a new background, that is a separate job. See how to remove the background from AI-generated videos for the workflow.

Where BGRemover.video fits

BGRemover.video removes and replaces backgrounds in both images and video, in the browser, and its homepage states exports have no watermark. Paid usage is credit-based; the price per credit is shown on the credits page when you are signed in, so we will not quote a figure here.

An honest framing: Craft and BGRemover.video are not substitutes. Craft makes a new image or layout. BGRemover.video takes footage or a photo you already have and isolates or replaces the subject. Many creators will use both: generate a background or template in Canva, then composite a real presenter onto it. See the video background removal feature page for the video side. If you are comparing dedicated tools, our tested best background removers roundup and the Photoroom comparison are the place to start.

What does this mean for sellers and product photos?

For e-commerce, Craft is useful for lifestyle backdrops and banner creative, but marketplace main-image rules are unchanged. Amazon, Etsy and Shopify still have their own requirements for primary images, and a generated scene is rarely acceptable as a main image where a clean white or neutral background is required. Our e-commerce background rules guide lists them, and how background removal improves product photo sales explains why clean cutouts still convert.

A sensible split: shoot or upload your real product photo, cut it out cleanly with a dedicated remover, then place it into whatever backdrop you generate. Keeping the product itself as a real photograph avoids the risk of a model subtly altering logos, labels or proportions, which matters if a listing must reflect the actual item. That caution is ours, not Canva's.

How do you try it today? (A practical checklist)

1. Open Dream Lab. In Canva, use the three-dot option in the left-hand menu and choose Dream Lab, per The Verge, and pick Canva Craft if the model selector offers it.
2. Prompt with structure. Name the elements you want and where they go. Craft's planner is built to handle placement and text, so describe layout, not just mood.
3. Choose the aspect ratio first. With 30+ ratios, generate at the final size rather than cropping later.
4. Convert to layers. Use Magic Layers to split the result so you can adjust a single element without regenerating.
5. Keep real subjects real. For people, products and footage, cut out the original with a dedicated tool and place it into the design.
6. For video, use a video tool. Run clips through a video background remover before importing them into your layout.
7. Check licensing and plan limits. Look at Canva's terms and your plan for commercial use and limits, since we could not confirm them.

What are people saying?

The most substantive early reaction we could find is the Product Hunt launch, which had 145 points and ranked fifth for the day when we checked. The hunter's comment praises legible text and the ability to edit pieces separately. Elsewhere on the Canva Product Hunt page, reviewers from earlier months complain about advanced controls, paywalled elements and uneven AI features. Those are about Canva overall, not Craft, so they are a reminder rather than a verdict. We did not find a Hacker News thread to cite.

Should you switch tools because of Canva Craft?

Not on the strength of one launch. If you already live in Canva, Craft is an addition inside a tool you use, with limits we could not confirm. If you do dedicated cutout work, especially video, nothing here changes your toolkit.

Watch three things over the next weeks: the Craft-fast and Craft-ultrafast releases, API pricing once it is public, and independent tests of how cleanly layers split real photographs. Until then, treat claims about "editable everything" as promising and unverified.

Conclusion

Canva Craft is a credible, design-focused image model with a smart planning step, 2K output, 30+ aspect ratios and a link to layered editing. It is best read as Canva owning more of its own stack, with a real benefit for text-heavy graphics and quick compositions.

It is not a background remover and, per everything we reviewed, not a video tool. If your day involves cutting a presenter out of a recording, replacing a messy room, or preparing product shots, you still need software built for that, and you can combine it with whatever backdrop Craft generates. Try Craft in Dream Lab for layouts and text. For footage and photos you already have, try BGRemover.video: the homepage states exports carry no watermark, and paid usage is by credit. We will update this post when Canva publishes pricing, limits or video features.