Canva launches Craft, its first in-house AI model for image creation
1–2 minutes
Canva has unveiled Craft, its first internally developed artificial intelligence model designed to generate images. The new tool is available on Dream Lab, the company’s testing platform, and was announced in the corporate newsroom on October 22.
How Canva Craft works
The system breaks the process into two stages: first, it organizes the visual composition, and then it generates the image. According to the company, the goal is to improve object positioning and text readability.
Generated images can be converted into editable elements using the Magic Layers feature. Canva also revealed that future versions, named Craft-fast and Craft-ultrafast, will be released later.
Why Canva invests in proprietary models
The research lab was established through the integration of teams and acquisitions made in recent years. The company acquired Kaleido in 2021 and Leonardo in 2024, bringing professionals together under the Canva Research umbrella in 2025.
The decision reflects a balance between quality and cost concerns. Since Canva processes over one billion designs per month, even small differences of a few cents per generation become significant. The strategy involves combining internally developed models with partner solutions, depending on the needs of each tool.
In an interview in April, co-founder Cliff Obrecht said that Canva planned to gather results from the new artificial intelligence tools before going public.
Frequently Asked Questions
Where can I use Canva Craft?
The model can be accessed directly in Canva’s Dream Lab. The company also announced that developers will have access via API.
How does the model organize image generation?
According to Canva, the system plans the composition before generating the image, aiming to improve object positioning and text readability.
Will Canva stop using models from other companies?
The approach combines internal development with partner technologies, depending on the specific needs of each tool.
