gen‑ai.news
← Back
Multimodal

Google says 180 billion images and videos now carry SynthID watermarks as detector goes public

Google says 180 billion images and videos now carry SynthID watermarks as detector goes public

Google has made its SynthID watermark detector publicly available, meaning anyone can now upload images, videos, or audio to check whether that content carries a SynthID watermark. Previously, the detection capability was limited to internal or partner use. The move marks a meaningful step toward giving ordinary users a concrete tool for identifying at least some AI-generated media.

The scale of deployment is notable. Google says more than 180 billion pieces of content already carry SynthID watermarks, a figure that reflects how widely the technology has been embedded across Google's own generative AI products as well as those of partners including OpenAI and Nvidia. SynthID works by embedding an imperceptible signal into the pixels of an image, the frames of a video, or the waveform of an audio file - a signal that persists even after common edits like compression, cropping, or color adjustment.

There is an important limitation to understand: the public detector identifies only content that carries a SynthID watermark. It cannot determine, in a general sense, whether a given piece of media was produced by AI. If an image was generated by a model that does not embed SynthID - whether a competing commercial tool or an open-source model - the detector will return no result. This means the tool is better understood as a provenance checker for a specific ecosystem rather than a broad AI-detection solution.

The broader context here is a growing industry effort to attach verifiable origin information to synthetic media. Initiatives like the Coalition for Content Provenance and Authenticity (C2PA) pursue a similar goal through cryptographic metadata rather than perceptual watermarking. SynthID's approach is complementary but distinct, and making the detector public at least lowers the barrier for anyone curious about content coming out of Google's AI pipeline and its partners. How widely consumers and journalists will actually use it remains to be seen.

Enjoy this story? Get the next one in your inbox.

Twice a week: the most important stories in generative image and video AI, distilled into a 2-minute read.

Free. Unsubscribe any time. No spam, ever.

Your next read

Someone Rebuilt Free, Open-Source Versions of Photoshop, Premiere, and Lightroom Using AI
Multimodal

Someone Rebuilt Free, Open-Source Versions of Photoshop, Premiere, and Lightroom Using AI

Software engineer Brandon Thomas has released ArtCraft, a free, open-source suite of applications designed to mirror much of Adobe Creative Cloud's core lineup - including equivalents of Photoshop, Premiere, Lightroom, After Effects, and more. The suite runs on macOS, Windows, and Linux, and was built with significant AI assistance. It represents one of the more ambitious open-source challenges to Adobe's dominance in creative software.

No image
Multimodal

Google’s new SynthID website can identify AI-generated media

Google has launched a public-facing website for SynthID, its AI watermarking technology, allowing anyone to upload and check whether an image, video, or audio clip was generated by AI. The tool extends SynthID beyond its previous developer and enterprise integrations, putting detection capabilities directly in the hands of everyday users. This marks a notable step toward accessible, practical tools for identifying synthetic media in the wild.

Reka AI's omni-model Rho-1 handles text, images, video, and robot control in a single model
Multimodal

Reka AI's omni-model Rho-1 handles text, images, video, and robot control in a single model

Reka AI has released Rho-1, a 19-billion-parameter model that handles text, images, video, and robot control actions within a single neural network. Rather than routing different tasks to specialized subsystems, it processes all modalities as tokens in one shared context window. The model was trained on 320 H100 GPUs over roughly three months, using considerably less compute than many leading models today.