SynthID Decoder
OpenAI's GPT-Image and ChatGPT, and Google's Gemini (Nano Banana), hide an invisible SynthID watermark in every picture they make. Upload an image and we decode whether it is there, in seconds, with no account and no API key.
Upload an image to decode its SynthID
Free, no account neededPaste or drag and drop it here, or upload a file
PNG works best • JPG, WebP • up to 50MB
Ctrl+Vto paste a copied image
Your image is written to a temporary file on our server in Germany while it is decoded, then deleted. We do not keep it or use it for training.
What is SynthID?
SynthID is Google DeepMind's invisible watermarking technology. Instead of a visible logo or a metadata tag that is lost on the first screenshot, it hides the mark in the pixels themselves, far below the level a human can see. In 2026 OpenAI adopted it for image generation: every output of GPT-Image, whether it comes from the API or from ChatGPT, carries the mark.
Each vendor uses its own secret key, so the OpenAI mark is different from the one Google embeds in Gemini output. This decoder checks for both without the keys, and makes no call to OpenAI or Google.
How the decoder works
1. Read the pixels
The decoder works on the image itself, not on its metadata or file tags, so it still works on pictures that have been stripped of their credentials.
2. Check both vendors
OpenAI and Google each use their own SynthID watermark. We run a separate detector for each, so you learn which vendor's mark was found.
3. Survive real-world sharing
Pictures are rarely passed around untouched. The decoder is built to find the mark after JPEG compression, downscaling, screenshots, padding and moderate crops, and Google's mark after paste, crop and most screenshots.
4. Tested against clean images
We test the decoder on images from other sources: camera photos, paintings and the outputs of other generators. The OpenAI decoder flagged none of more than 4,000 such images; the Google detector flagged none of more than 22,000 held-out clean test images.
What a negative result means, and what it does not
A detection is strong evidence the image was produced by an OpenAI or Google model. A non-detection is not proof that it is real: it only says neither mark was found. Images from Midjourney, Flux, Stable Diffusion, Qwen and every other generator return negative here by design. For those, use the full Sightova AI image detector, which does not rely on a watermark.
The OpenAI mark survives JPEG compression, downscaling, screenshots, padding and moderate crops. It can be lost when an image is heavily zoomed or stretched, or when the image is too plain and smooth to carry it.
Google's mark now survives paste, crop and most screenshots. A heavy JPEG share of a screenshot is the weakest case. A negative on a heavily recompressed Gemini image still means little: check it with the full detector.
Need this in your own pipeline?
Watermarks only cover the vendors that embed them. Sightova's AI-image model and document-tampering check look at the pixels themselves, through the dashboard or the REST API.