SynthID watermark
An invisible signal in the pixels. It is built to survive screenshots, crops, filters and compression, and both Google’s SynthID Detector and OpenAI’s checker read it.
ChatGPT · OpenAI API · Codex images
Since May 2026, every image from ChatGPT, the OpenAI API and Codex carries Google’s SynthID watermark and a C2PA Content Credentials manifest. Spoofzy removes both in one pass and hands back the file at the size you generated it, in 10–20 seconds.
chatgpt-image.png
What changed in May 2026
OpenAI added Google DeepMind’s SynthID watermark next to the C2PA Content Credentials it already wrote, so an image stays traceable even after its metadata is gone.
An invisible signal in the pixels. It is built to survive screenshots, crops, filters and compression, and both Google’s SynthID Detector and OpenAI’s checker read it.
A signed manifest in the file that says OpenAI made the image. Instagram, TikTok and LinkedIn read it when they decide whether to show an AI label.
Screenshots and re-saves drop the C2PA manifest but leave SynthID in the pixels, so the image still checks as OpenAI-made.
Spoofzy removes the SynthID signal and strips C2PA, IPTC AI tags and generator EXIF together, at the size ChatGPT gave you.
How it works for you
One image or a whole session of renders: the same pass runs in Spoofzy Cloud, in Telegram and through the API.
One image or a whole batch, in Spoofzy Cloud, Telegram or through the API.
SynthID removed, C2PA stripped, pixel size and colour kept.
Upload the original and the clean copy to openai.com/verify or Google’s SynthID Detector.
Send the clean file on, or turn it into unique copies with the Video & Image Spoofer.
Questions
Yes. OpenAI said in May 2026 that images made with ChatGPT, the OpenAI API and Codex carry Google DeepMind’s SynthID watermark alongside C2PA Content Credentials.
No. Stripping metadata, screenshotting or re-saving removes the C2PA manifest but leaves SynthID in the pixels. Spoofzy removes both in the same pass.
Upload the original and the clean copy separately to OpenAI’s checker at openai.com/verify, which reads SynthID and C2PA on OpenAI images, or to Google’s SynthID Detector at synthid.com. Gemini’s built-in check only reads Google’s own watermark, so it cannot confirm a ChatGPT image either way.
No. Spoofzy returns the file at the resolution you uploaded, up to native 4K and beyond, instead of a reduced preview size.
Yes. Drop a batch into Spoofzy Cloud, hand it to your team in Telegram, or call the REST API with presigned URLs and signed webhooks.
SynthID Remover API and MCPYes. The same SynthID Remover cleans Gemini, Nano Banana and Imagen renders, so one workflow covers every model your team uses.
SynthID RemoverYes. New accounts get 40 free Credits with no card, enough for 2 clean images at 16 Credits each.
Clean your first 2 images free. No card, no GPU setup.