SynthID bypass: screenshots, crops and compression
See how screenshots, crops and compression affect SynthID, then check your processed image with Gemini.
SynthID is designed to survive common image edits. Screenshots, crops and compressed copies can still carry it. Use Gemini to check the original and the processed image, then compare the responses.
Check the watermark in the edited image
SynthID is embedded in the image itself. Cropping a corner, changing dimensions or saving a smaller copy can leave the watermark intact.
To check a change, upload the original and the edited copy to Gemini separately. Compare the SynthID responses alongside the image quality.
Does screenshotting remove SynthID?
A screenshot can still carry detectable SynthID. Google’s verification instructions accept screenshots and recommend framing the image tightly.
Keep the original download for comparison. A screenshot reflects the size and quality displayed on screen, which can affect fine detail.
What changes when you crop or compress an image?
Cropping changes the composition; compression reduces file size and can soften detail. SynthID is designed to remain detectable through both kinds of edit.
Check the saved image in Gemini after processing. The table separates the visible change from the watermark check.
| Edit | Visible effect | SynthID check |
|---|---|---|
| Crop the picture | Changes framing or removes a visible corner. | Check the cropped file in Gemini. |
| Compress or resize | Reduces file size or dimensions; detail may soften. | Check the saved copy in Gemini. |
| Turn off Gemini’s visible watermark | Removes the visible mark controlled by that setting. | SynthID and Content Credentials remain enabled. |
Source: Google: visible watermark settings leave SynthID enabled
Read the Gemini response
“SynthID detected” means Gemini found a watermark. “Not detected” means it found none in that file. An inconclusive response needs a closer look at the input.
Gemini’s SynthID check covers Google AI. Other AI-generated images can also return a “not detected” result.
Compare demonstrations fairly
A useful demonstration pairs the original image with the processed copy and both verification responses. It also lets you inspect faces, text and fine detail.
When a tool publishes a success rate, look for the sample size, detector and treatment of failed results. The SynthID-Image research paper shows how defined test conditions make results easier to compare.
Source: Google researchers: SynthID-Image robustness evaluation
Run a before-and-after check
Keep the original, processed copy and Gemini responses together. A simple comparison is easy to repeat and share with your team.
- Check the original image in Gemini.
- Process a separate copy and download it.
- Check that copy with the same verifier.
- Save both responses with the matching files.
- Compare faces, text, colour and framing at full size.
Source: Our guide to checking a supported image before and after processing
Try Spoofzy’s AI Watermark Processing
Open AI Watermark Processing in Telegram, send a supported Google AI image and download the processed copy. The product page shows current image support and a real demonstration.
Use the steps above to compare your result in Gemini. For GPS and camera details, choose Metadata Privacy; visible logos need a visual-editing tool. Keep any required AI disclosure with the finished post.
Source: AI Watermark Processing: current image workflow and limits
Put your Google AI images to work.
Try AI Watermark Processing in Telegram and check your result in Gemini.
Explore AI Watermark Processing