This section covers two capabilities that are easy to conflate but answer different questions:
- Visual Verification — public, no-auth endpoints that check whether an image or video carries Bria's provenance markings: a C2PA manifest and an invisible watermark. Use this when the question is "was this generated by Bria?" — checking your own output, a user's upload, or a file received from someone else.
- Attribution — how Bria attributes what a model generates back to the licensed training data that shaped it, and what to run when a Bria model is hosted outside Bria's cloud (BYOC, on-premises, self-hosted weights). This is what stands behind Bria's commercial indemnity and the revenue-share paid to data partners.
If you're checking authenticity, start with Visual Verification. If you're asking how data owners get compensated, or what you need to run when self-hosting a Bria model, start with Attribution.