Original AI creative models

Canva Craft

Canva's own image model, up to 2K, in Dream Lab.

Launched
Visit website
Canva Craft — official product preview
Official product imageryView full image ↗

What it does

Canva Craft is the first image model trained by Canva Research. The newsroom describes two stages: one plans the picture, including what appears, where it sits, and any text, and the next renders it at up to 2K in more than 30 aspect ratios, or a custom ratio. Prompts can be written in any language, and text in the image can be in 55 languages. It is available to Canva users in Dream Lab, and on the Connect API as the model value craft_v1_2k. The public training summary names this version Craft-1 2K. Magic Layers can turn an output into editable design elements. Canva Shield covers input and output checks.

Where it fits

Open Dream Lab, describe the image, and choose an aspect ratio. Check faces, text, and the layout before you drop the picture into a design. If you need pieces you can still move, run Magic Layers and review each one.

Before you commit

The newsroom says Craft is available to all Canva users and does not list a separate price, so check the current Canva plan for generation limits. Craft-fast and Craft-ultrafast were described as upcoming, not as this release. The model can render people and text; review both before you publish. The live newsroom page returned a Cloudflare check from this network, so the preview is the official output collage from that article. This check did not generate an image.

Release & update notes

Dates below refer to the linked announcement or Product Hunt launch. They are separate from the date this profile was added.

Sources & editorial notes

This profile summarizes public product information and adds workflow guidance. It is not a hands-on performance review. Check the official website for current prices, limits, and availability.

Source check:

Keep exploring

0/3 Compare tools