Reverse engineering Gemini's SynthID detection
_tk_
129 points
48 comments
April 09, 2026
Related Discussions
Found 5 related stories in 95.2ms across 8,358 title embeddings via pgvector HNSW
- Reverse engineering circuitry in a Spacelab computer from 1980 elpocko · 96 pts · May 23, 2026 · 54% similar
- Gemini 3.5 Flash spectraldrift · 697 pts · May 19, 2026 · 49% similar
- Gemini 3.1 Flash Live: Making audio AI more natural and reliable meetpateltech · 12 pts · March 26, 2026 · 49% similar
- Reverse engineering Android malware from popular Chinese projectors 3abiton · 19 pts · May 18, 2026 · 48% similar
- Gemini 3.5 Flash: frontier intelligence with action meetpateltech · 179 pts · May 19, 2026 · 48% similar
Discussion Highlights (11 comments)
andrewmcwatters
> We're actively collecting pure black and pure white images generated by Nano Banana Pro to improve multi-resolution watermark extraction. Oh hey, neat. I mentioned this specific method of extracting SynthID a while back.[1] Glad to see someone take it up. [1]: https://news.ycombinator.com/item?id=47169146#47169767
kelsey98765431
if you downscale then upscale it removes the watermark
armanj
kinda ironic you can clearly see signs of Claude, as it shows misaligning table walls in the readme doc
khernandezrt
Ok i get that eventually someone was gonna do this but why would we want to purposely remove one of the only ways of detecting if an image is ai generated or not...?
M4v3R
SynthID is visible in some generations (areas with a lot of edges, or text), I wonder if this would make them look better.
sodacanner
I don't understand all the handwringing. If it's this easy to remove SynthID from an AI-generated image then it wasn't a good solution in the first place.
doctorpangloss
Okay... this tests its own ability to remove the watermark against its own detector. It doesn't test against Gemini's SynthID app. So it does nothing...
Tiberium
Seems like a very low-quality AI-assisted research repo, and it doesn't even properly test against Google's own SynthID detector. It's not hard at all (with some LLM assistance, for example) to reverse-engineer network requests to be able to do SynthID detection without a browser instance or Gemini access, and then you'd have a ground truth.
coppsilgold
Inserting an undetectable 1-bit watermark into a multi megapixel image is not particularly difficult. If you assume competence from Google, they probably have two different watermarks. A sloppy one they offer an online oracle for and one they keep in reserve for themselves (and law enforcement requests). Also given that it's Google we are dealing with here, they probably save every single image generated (or at least its neural hash) and tie it to your account in their database.
m00dy
It's not really hard as it looks https://deepwalker.xyz/blog/bypassing-synthid-in-gemini-phot...
stevomacdaddy
Im confident i saw the watermark in use today, in nano banana, i copied the image from chrome into slack. the resulting upload was a black square with a red dot. and not the image i had generated.