Google opens the SynthID Detector to anyone, in English

Google says the SynthID Detector is now open to anyone. In a 7 October post, Pushmeet Kohli, vice president for science and strategic initiatives at Google DeepMind, writes that the tool is “available globally in English starting today.”

SynthID is an invisible watermark Google embeds in AI output. Kohli writes that since the 2023 launch, Google has used “imperceptible watermarks across images, video, and audio.” The detector looks for that mark. It is a check for a watermark, which is a different task from judging whether a picture looks real.

What Google says you can check

Kohli writes that anyone can check whether an image, a video, or an audio file “was made with AI from Google or our partners, including OpenAI, NVIDIA, Kakao, and soon, Apple.” Those four names, with Apple marked “soon,” are on Google’s page. The post does not say Apple Intelligence by name, and it does not give a date for Apple.

The Verge says anyone can use the tool by logging in at synthid.com. That is the site Google’s post links to.

TechCrunch lists the formats it says the site accepts. For images: JPG, JPEG, PNG, BMP, WEBP, AVIF, HEIC, HEIF, TIFF, TIF, and GIF. For video: MP4, MOV, and WEBM. For audio: WAV, MP3, OGG, FLAC, AAC, and M4A. Google’s post itself names only the media types: images, video, and audio.

TechCrunch also writes that Google’s own generators, including Nano Banana, Veo, and Lyria, plus Gemini, Flow, ProducerAI, and Vids, watermark with SynthID.

Who had it before today

Kohli writes that last year Google introduced an early version of the detector “to help media professionals verify AI-generated content,” and that access is now expanding to everyone. The Verge says the earlier tool was for media professionals and that the public step is new. TechCrunch says the earlier testers included journalists, media professionals, and researchers, and that the trial was shown at Google I/O last year.

Google’s post also says built-in checks in Search, the Gemini app, and Chrome “now regularly handle over 1 million requests daily.” The watermark totals on the same page are Google’s: “over 180 billion images and videos, along with 240,000 years of audio content.” Both figures are the company’s.

What a missing mark means

A watermark detector can report a mark it knows how to read. If a file has no SynthID mark, that result does not show the file was made by a person. A tool that never embeds SynthID, or a file whose mark was stripped, can still be synthetic. 9to5Google makes the same point in its own words: detection works when the generator inserted the watermark to begin with. TechCrunch adds that watermark checks are fallible and often miss content from the maker’s own models, pointing at other companies’ systems as well as this kind of tool.

Google describes the detector as a way to check whether a file was made with AI from Google or the named partners, so that partner list is the tool’s scope.

What to do with a result

For a developer or an editor, the practical use is narrow and useful. Upload an image, a video, or an audio file and see whether Google reports a SynthID watermark from Google, OpenAI, NVIDIA, or Kakao. Treat “soon, Apple” as Google’s timing, not as support that is live today. Treat a clean result as “no SynthID mark found,” and keep a separate check if the question is whether a person made the file.