Guide/Instagram
Can you remove SynthID from an image?
Updated
No. SynthID is an invisible watermark that Google DeepMind builds into the content itself (the pixels of an image, the frames of a video, the samples of an audio clip), so no metadata tool can remove it, and ScrubPhoto does not try. What a metadata tool can remove is the separate C2PA Content Credentials record that some Google images carry next to the watermark.
What is SynthID
Google DeepMind's SynthID page calls it a watermarking tool designed for AI-generated content. It says the watermark is embedded directly into AI-generated images, audio, text or video, and that people cannot perceive it while SynthID's own technology can detect it.
The clearest explanation is in DeepMind's August 2023 launch post. It says the image watermark is embedded in the pixels, and it contrasts that with metadata, which can be removed by hand or lost when a file is edited. A pixel watermark, the post explains, stays detectable even when the metadata is gone.
Google's Gemini help page adds that the watermark usually survives everyday changes to a file, like compression. That describes how it was built, not an invitation to test it.
Which Google tools add SynthID
Google lists several products on its own pages:
- Imagen. The 2023 launch post introduced SynthID for images produced by Imagen.
- Image generation in the Gemini app. Google DeepMind's page for Nano Banana Pro says all images from image generation in Gemini are watermarked with SynthID.
- Veo. DeepMind's Veo page says videos made with Veo will be marked with SynthID.
- Audio from Lyria and the podcast feature of NotebookLM, plus text from the Gemini app, according to the SynthID page.
Google said in November 2025 that more than 20 billion pieces of AI-generated content had been watermarked with SynthID since 2023. The SynthID page also says its Detector can check files made with AI from Google or its partners, and it names OpenAI, NVIDIA and Kakao, with Apple listed as coming soon.
Remove C2PA and SynthID: two layers, two answers
A Gemini image can carry two separate marks. Google's November 2025 post says images generated by Nano Banana Pro in the Gemini app, Vertex AI and Google Ads have C2PA metadata embedded. That is a Content Credentials record stored in the file's metadata, next to the SynthID watermark in the pixels.
The two layers behave differently. Content Credentials are a block of bytes beside the image data, so they can be lifted out without changing a single pixel. SynthID is part of the image data itself. Removing it would mean changing the picture, and a tool that changes the picture is not a metadata tool.
Gemini's help page puts it the same way: watermarks are embedded directly into content, while metadata adds context alongside the original file. That split decides what ScrubPhoto can touch.
What ScrubPhoto can and cannot do with a Gemini image
It can read and remove the metadata layer. The checker shows whether a file carries C2PA and which app signed it, the IPTC digital source type (created by AI, edited with AI, and so on), and the XMP creator tool. The home tool then removes C2PA, XMP, IPTC and EXIF from JPEG, PNG and WebP files without decoding the image.
It cannot detect SynthID, and it cannot remove it. Because the scrub is lossless, the pixels in the output are byte-for-byte identical to the original, which also means the watermark is exactly where it was. Even when the checker reports no metadata trigger, it reminds you that invisible watermarks are outside what it can see.
Meta's own AI images follow the same pattern. Meta says they carry an invisible watermark plus IPTC metadata, and removing the metadata does not touch the watermark. The Gemini page and the limits page list everything this tool does not do.
Does Instagram read SynthID?
Meta's published approach to its AI info label (Instagram has also shown it as AI content, and first called it Made with AI) says Meta looks for industry standard AI indicators in the file, meaning C2PA Content Credentials and IPTC metadata, or a disclosure from the person posting. The pages we checked do not say whether Meta also reads SynthID, so do not assume either way.
What you can know is what your file carries. If a Gemini export shows C2PA in the checker, the Instagram page explains how Meta tends to react, and the guide on why an Instagram post says AI info covers the label itself.
Meta also requires people to disclose photorealistic video and realistic audio that was digitally created or altered. Removing metadata does not change that duty.
SynthID detector options from Google
If you want to know whether an image or clip carries SynthID, use Google's own tools. Nothing on this site can answer that question.
- The Gemini app. Google's November 2025 post says you can upload an image and ask a question like "Was this created with Google AI?" or "Is this AI-generated?", and Gemini checks for the watermark. The Gemini help page says the check now covers images, videos and audio, one file at a time.
- The SynthID Detector, a verification portal. Google says it launched in 2025 for journalists and media professionals and that access is expanding to everyone.
- Search and Chrome. The SynthID page says Gemini, Search and Chrome will check for the watermark and tell you if they find one.
Read the result carefully. Google's help page says a detected watermark means all or part of the content was made or edited by Google's AI. No watermark means Google AI did not make it, but another AI system still could have.
Why this site will not offer a SynthID remover
People search for "synthid remover" and "remove ai watermark" every month. This page will not answer them with ways to damage an image, for two reasons. A watermark is part of the picture, so any method that weakens it also changes the photo you meant to post. And the watermark exists so people can tell what made an image, which is not something a privacy tool should work against.
ScrubPhoto is built for metadata: location, serial numbers, editor tags, and labels that landed on real photos by mistake. It removes data and never adds any. It will not write fake camera details into an AI image, because inventing provenance is a different act from removing it. For the metadata side, start with what C2PA Content Credentials are.
Questions about removing SynthID
Can you remove SynthID from an image?
No. SynthID is embedded in the image data itself, not stored in the metadata, so removing metadata leaves it in place. ScrubPhoto neither detects nor removes it, and this site does not publish ways to defeat it.
Does removing C2PA metadata remove SynthID?
No, because they are separate layers. C2PA Content Credentials are a metadata block that can be removed without touching the pixels, while SynthID lives in the pixels, so a lossless metadata scrub keeps the watermark exactly as it was.
What is SynthID used for?
Google uses it to mark content made or edited with its AI models so that it can be identified later. Google's pages name Imagen, image generation in the Gemini app, Veo, Lyria audio and Gemini text as covered.
How can I check if an image has SynthID?
Upload it to the Gemini app and ask whether Google AI created or edited it, or use Google's SynthID Detector. Google says the Detector also checks content from partners it names, including OpenAI, NVIDIA and Kakao.
Is SynthID the same as the AI label on Instagram?
No. The AI info label is Meta's label, and Meta says it is driven by C2PA and IPTC metadata or by your own disclosure. SynthID is Google's watermark. A Gemini image can carry both a SynthID watermark and C2PA metadata.
Does ScrubPhoto change the pixels of my image?
Not for JPEG, PNG or WebP. Those are scrubbed without decoding, and the result shows "Pixels identical". HEIC is the exception: in Safari it is re-saved as a JPEG and marked "re-encoded, not lossless".
Sources
Official pages this page was checked against in October 2026: