ppl.studio

What is SynthID?

SynthID is Google DeepMind's invisible watermarking system for AI-generated content. It embeds an imperceptible signal directly into the pixel values of an image (or audio waveform, or text token distribution) that can be detected by Google's verifier even after common transformations — JPEG compression, color shifts, cropping, screenshot capture, and re-saving. Unlike visible watermarks, SynthID does not change the visual appearance of the image. It is the technical companion to C2PA: where C2PA carries signed metadata that can be stripped, SynthID survives metadata removal because the watermark is in the image itself. Google's Imagen and Veo models embed SynthID by default; Meta and Microsoft have signaled support. For brands shipping AI content, SynthID is becoming relevant because platforms are beginning to use SynthID detection (alongside C2PA reads) to auto-apply 'Made with AI' labels — meaning your content can be auto-labeled even if you didn't disclose it manually.

Key statistics

  • SynthID survives JPEG compression, cropping, color shifts, and screenshot recapture (Google DeepMind technical paper, 2024).
  • Google's Imagen 3 and Veo models embed SynthID by default on all generations (Google AI announcements 2024–2025).
  • Meta, TikTok, and YouTube now read both C2PA and SynthID signals to auto-apply AI labels (platform updates, 2025).
See it in action — create UGC

Related terms

Back to glossary