Canva introduces ‘Magic Layers’, allowing you to convert an image into editable layers. This should make it easier to adjust existing visuals without having to rebuild the design from scratch.
Canva launches Magic Layers, an AI feature that converts flat images into editable designs with separate layers. The technology recognizes text, objects, and layout structures in a flat image and reconstructs them as individual elements that can be customized. Magic Layers is currently only available in public beta in the United States, the United Kingdom, Canada, and Australia. Global availability will follow, according to Canva.
Recognizing structures
With Magic Layers, Canva aims to address a limitation of many AI-generated visuals. Images created via generative AI often end up as static files such as PNG or JPEG. Once exported, it is difficult to modify individual parts without starting over.
The new feature analyzes an entire design and reconstructs its structure. Text becomes recognizable again as editable text boxes. Objects appear as separate elements that users can move or adjust. Backgrounds remain separate from foreground objects.
Editing without starting over
The feature doesn’t just work with designs created within Canva itself. Users can also import external images or AI-generated visuals and have them converted into a layered design within the editor.
After conversion, components can be adjusted individually. Users can move or resize objects, change fonts, and adjust colors. The layout can also be recomposed. The result can then be used in presentations, videos, or documents.
Underlying AI model
Magic Layers is built on the Canva Design Model, an AI model that the company says is designed to generate editable designs. The model also forms the basis for features in Canva AI and integrations with platforms such as ChatGPT, Claude, and Microsoft Copilot.
Magic Layers is currently being rolled out in public beta in the United States, the United Kingdom, Canada, and Australia. Broader international availability will follow later.