Google Makes Gemini Media Watermarks Optional
Google announced on August 14 that it is rolling out a setting to remove visible watermarks from AI-generated images, videos, and music in Gemini and Flow. According to TechCrunch and The Verge, the change applies to Nano Banana, Omni, and Lyria outputs, while invisible SynthID watermarks and C2PA metadata remain embedded for provenance checks.
Google announced on August 14 that users can disable a visible "Media watermark" on AI-generated images, videos, and music made through Gemini and Flow, its AI video-generation tool. TechCrunch reports that the setting is rolling out in the coming days for output from Google's Nano Banana, Omni, and Lyria models.
The optional mark is Google's visible "sparkle" indicator, which The Verge reports appears in the bottom-right corner of generated media. Users can find the toggle under Settings > Media Watermark in Gemini and Flow. TechCrunch reported that support in Google Search is coming soon, citing Josh Woodward, Google's vice president for Gemini.
Invisible provenance markers remain
Disabling the visible marker does not remove Google's other labeling mechanisms. Both TechCrunch and The Verge report that generated content will retain invisible SynthID watermarks and C2PA-related metadata. SynthID is Google's watermarking system for identifying AI-generated media, while C2PA is a technical standard for recording content provenance and edits.
Woodward described the policy tradeoff in a post cited by TechCrunch: "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."
The Verge reported that Gemini or Google Search can be used to check whether an image carries these markers. The source did not provide comparable verification details for video or audio in the retrieved report.
Implications for media workflows
The change removes a persistent visual artifact that can limit use of generated media in professional or creative assets, while retaining machine-readable provenance signals. The Verge notes that users had already been editing visible watermarks out of generated content, and that OpenAI and Meta generally rely on non-visible signals rather than visible marks.
For practitioners building review or moderation workflows, the distinction matters
visible labels support rapid human inspection, whereas SynthID and C2PA checks require compatible detection or metadata-reading paths. Across the industry, optional visible labels can make provenance verification more dependent on platform tooling and downstream support for standards such as C2PA.
Key Points
- 1Google is rolling out an option to disable visible labels on Gemini and Flow media generated by Nano Banana, Omni, and Lyria.
- 2Invisible SynthID watermarks and C2PA metadata remain after the visible mark is disabled, preserving a provenance-checking mechanism.
- 3Across AI media workflows, optional visual labels increase reliance on detection tools and metadata support rather than immediate human inspection.
Scoring Rationale
The update affects provenance labeling for a major consumer AI media platform and is relevant to teams designing media-review, moderation, and authenticity workflows. Its practical effect is narrower than a model release or a change to watermarking standards because invisible provenance mechanisms remain in place.
Sources
Public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
