SynthID is Google DeepMind's tool for putting a hidden watermark into AI-made images, video, audio and text, and for finding that mark later.
In plain words
When Google's AI tools make a picture, video, sound clip or text, SynthID hides a signal in it that people cannot notice but its software can find. The mark does not lower image or video quality, and audio marks survive added noise or MP3 compression. For text, it shifts the scores a model uses to pick each next word (token). Gemini, Search and Chrome can check files for the mark, and the SynthID Detector now lets anyone check images, video and audio. Sample text-mark code is on GitHub.
A quick example
Imagine a teacher is sent a dramatic photo of a storm and wants to know where it came from. They upload it to a SynthID checker, which reports whether the hidden mark from a supported AI tool is inside the picture.
Where you'll meet it
- Labelling images, video, audio and text made with Google's AI tools.
- Checking whether a file came from AI tools by Google or its partners.
Related
Sources
- DocsSynthID, Google DeepMind · read 13 Nov 2026
- DocsSynthID Text, Google DeepMind (GitHub) · read 13 Nov 2026
- DocsWe're making it easier to identify AI-generated content globally, Google · read 13 Nov 2026
Checked by WhatIsAI against the sources above · 13 Nov 2026. Spotted an error? Report it.