Claude can now generate web pages that place real HTML elements inside canvas graphics, combining elaborate visual effects with text and controls that remain interactive. The feature lets developers bend or position web content in ways normally associated with a flat canvas image without giving up selectable text, links or buttons.
A demonstration includes 15 designs, such as pages rendered as knitted patterns, three-dimensional cubes and exploded layers. Because the content remains HTML rather than pixels baked into an image, visitors can highlight and copy text. That also gives the approach a path to better accessibility than conventional canvas-only interfaces, which browsers often treat as a single graphic.
Google is developing HTML-in-Canvas as a web standard. Current support is limited to Chrome and Edge version 157 or newer, along with Claude desktop apps from version 5.11492. Safari and Firefox do not support it yet, so pages that depend on the technique will need fallbacks.
Users must also ask Claude for HTML-in-Canvas explicitly in their prompt. The release is therefore more of a new construction option than a universal change to generated websites, but it gives designers a way to create unusual interfaces without turning their content into an inaccessible picture.