Overview
- On Friday, August 14, 2026, Google added a “Media watermark” toggle in the Gemini app and Flow that lets users remove the visible corner sparkle on images, video, and music generated by Nano Banana, Omni, and Lyria models.
- Turning off the visible mark does not remove backend signals because every Gemini-generated file will still carry invisible SynthID watermarks and C2PA Content Credentials that machines can read and Gemini can surface.
- Gemini now verifies SynthID in images, audio, and video and can scan videos across visual and audio tracks to flag the exact segments that contain Google AI‑generated material.
- The rollout is phased and restricted in some cases: Google will keep visible labels where local laws require them, work and school accounts may lack the toggle, and users in India, South Korea, and Vietnam may need an AI Ultra subscription.
- SynthID is designed to survive common edits but independent tests show detection can weaken after repeated transforms and some third‑party services claim ways to bypass it, so experts say these signals should be used as one layer in broader provenance and review policies.