Google’s new SynthID website can identify AI-generated media
TechCrunch Ivan Mehta ● Covered by 3 sources
Google opened a site that checks if an image, video, or audio clip was made by AI. It uses SynthID, and Google says people already make 1 million checks a day.
Based on reporting by TechCrunch, Ivan Mehta — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Google has opened SynthID verification to everyone, turning a tool that was previously limited to selected journalists, media professionals, and researchers into a public website. The service checks images, video clips, and audio files for signs that they were generated with AI.
The site accepts a wide spread of file types. For images, that includes JPG, JPEG, PNG, BMP, WEBP, AVIF, HEIC, HEIF, TIFF, TIF, and GIF. For video, it works with MP4, MOV, and WEBM. For audio, it supports WAV, MP3, OGG, FLAC, AAC, and M4A.
Google is leaning on SynthID, the watermarking system it introduced in 2023. Its own models and products — including Nano Banana, Veo, Lyria, Gemini, Flow, ProducerAI, and Vids — use SynthID to watermark generated media. The company has also built verification into the Gemini app and Google Chrome.
The pitch is simple: make it easier to check what’s real and what isn’t. But the limits are already familiar. Google says people make 1 million verification requests every day, yet it also acknowledges these tools are not perfect and can miss content made by the same companies’ own models. OpenAI, Nvidia, and Kakao support SynthID too, while OpenAI runs its own checker. Apple is said to be adding support soon.
My take — AI-written commentary, not fact-checked reporting
This is the right direction, mostly because the alternative is shrugging at a web full of synthetic mush and hoping for the best. But watermarking only matters if the tools keep working when the content gets messy, remixed, and reposted — which is exactly where these systems tend to get a bit embarrassed.
Read more about this at: TechCrunch
Related stories
Google's SynthID watermark is hard to break, but it doesn't solve AI misinformation
Ars Technica · 2 months ago ·
46