Skip to main content
A try-on result is a real photo of a real person that a generative model has altered. Under Article 50 of the EU AI Act, that has to be disclosed. Every result we return is marked in two ways.

​
What gets applied

A visible badge. An AI MODIFIED pill is composited into the bottom-right corner of the result, based on the European Commission’s official icon. It is the “partially AI-modified” variant, which is the correct one here: the shopper’s own photo is the starting point, not a wholly synthetic image. The badge switches between its black and white versions depending on how light that corner is, so it stays readable over any garment. Machine-readable provenance. Every result also carries XMP metadata with the IPTC DigitalSourceType set to trainedAlgorithmicMedia, the interoperable term platforms and detection tools look for. This is written on every generation and cannot be turned off.
The badge sits in the opposite corner from your brand watermark, so the two never overlap.

​
Turning the visible badge off

The metadata always stays. The badge is yours to control, in two places. Account default. Toggle EU AI content label on the Settings → Watermark & AI label page in the platform dashboard. It applies to every generation on your account, including the MCP server. Per call. Pass output: { aiLabel: false } on POST /try-on to drop it for a single generation, whatever your account default is.
Turning the badge off does not remove the obligation, it moves it. You become responsible for telling your end users that these images are AI-modified, by whatever means fits your product.

​
Genlook apps and storefront widgets

Results generated through the Shopify, WooCommerce, PrestaShop and Shopline apps always carry the badge. There is no merchant setting for it, and the API toggle does not reach them.

​
See also