Google adds a toggle in Gemini and Flow to remove visible watermarks from AI-generated images, videos, and music; SynthID watermarks and C2PA metadata remain

Google will still embed invisible SynthID and C2PA watermarks into AI-generated images, videos, and music.

The Verge 2026-08-14 Emma Roth

Context & Ripple Effects

Google introduced SynthID as an invisible marker for Imagen-made images, then expanded the system to AI-generated text. Its recent assessment that SynthID is difficult to remove also acknowledged that labels cannot cover every source of AI-generated content.

Today’s change separates presentation from provenance: Gemini and Flow users can remove an on-image, video, or audio marker while Google retains SynthID and C2PA data for machine-based identification. That distinction matters as Google has also made Gemini capable of detecting Google AI-created or edited images.

First-order effects

Gemini and Flow users can produce media without a visible watermark, while Google continues to attach invisible SynthID signals and C2PA metadata.
People evaluating media from those tools lose an immediate visual cue and must rely on compatible provenance or verification tools instead.

Second-order effects

Google’s verification features become more important for identifying its outputs because the visible label is now optional.
OpenAI’s earlier adoption of SynthID for AI images gives Google’s invisible-provenance approach relevance beyond its own generation tools, even as visible labeling becomes less consistent.

Third-order effects

Provenance systems are becoming a separate layer from the creative output itself: platforms can preserve origin data without requiring creators to display a platform-branded mark.
The move reinforces the limits of watermarking as a universal labeling regime; verification can identify participating systems but cannot label AI media made without such signals.

The trend: AI platforms are shifting from mandatory visible labels toward interoperable, machine-readable provenance that is less intrusive in creative workflows.

Discussion

@dotdotdotdash @dotdotdotdash on bluesky
Fakers will fake. They will see watermarks, or rather the lack of them, as an opportunity make their fakes seem genuine. www.bleepingcomputer.com/news/ securit...
@geminiapp @geminiapp on x
Rolling out over the next few days, you can decide if your image, video, and music creations made with Gemini will have visible watermarks. Invisible SynthID watermarks and C2PA metadata will always stay embedded in the background. Simply go to your settings and toggle “Show
@joshwoodward Josh Woodward on x
✅ Papercut fixed: You can now toggle visible watermarks on or off in Gemini and Flow, with Search coming next. This applies to watermarks on all images (Nano Banana), videos (Omni), and songs (Lyria) except in countries where it's required by law to keep them.
@joshwoodward Josh Woodward on x
We're striking a balance here between creative control and safety: while the visible watermarks are now optional, invisible SynthID watermarks and C2PA metadata are still being used for transparency. So you can still use Gemini or Search to see if an image was AI-generated.
@joshwoodward Josh Woodward on x
We're also open-sourcing Credentio, a new C++ library for C2PA Content Credentials to help developers build better provenance tools. Check it out: https://developers.googleblog.com/ ...

Chronicles