Is the SynthID Detector free?

Yes. Google's SynthID Detector at synthid.com is free but needs a sign-in and has a daily quota. The metadata checker on this page is free, has no limit and needs no account.

Does this site read the SynthID watermark itself?

No. Google has not released a public SynthID detection API, so no third-party site can read the watermark directly. This page reads Content Credentials and AI metadata on your device and sends you to Google's official detector for the watermark.

Can the SynthID Detector check ChatGPT images?

Yes. Images made with ChatGPT, Codex and the OpenAI API carry both SynthID and Content Credentials, and Google's detector recognizes them. The checker above also shows the OpenAI Content Credentials while they are still attached.

Can it detect Midjourney, Stable Diffusion or FLUX images?

Not through the watermark, because those tools do not use SynthID. The metadata check can still flag them when the file keeps its original generator tags, such as Midjourney job data or Stable Diffusion parameters.

Can SynthID detect AI-written text?

Google watermarks Gemini text with SynthID, but the public SynthID Detector only accepts images, video and audio. The open-source SynthID Text library can only detect text watermarked with your own keys, not Gemini's.

Do screenshots or compression remove SynthID?

SynthID is designed to survive common changes like compression, resizing, cropping and filters, though many stacked edits can defeat it. Screenshots and social uploads almost always strip Content Credentials, which is why the two checks complement each other.

Does "not detected" prove a file is real?

No. It only means no SynthID mark was found. The content may come from a model that does not use SynthID, or it may have been altered. Look at the metadata, the source and the context as well.

Are my files uploaded?

Not here. The checker reads the file inside your browser. If you then use Google's detector, Google processes the file in real time, deletes it after returning the result, and keeps a signature of it for 24 hours to enforce quotas.

Is this site affiliated with Google?

No. SynthID Detector at synthiddetector.app is an independent guide and checker. SynthID is a trademark of Google LLC.

SynthID Detector: Check images, video and audio for Google AI watermarks

Drop a file to read its Content Credentials and AI metadata in your browser. Nothing is uploaded. Then confirm the invisible SynthID watermark with Google's official detector.

Independent site. Not affiliated with Google or Google DeepMind.

Drop an image, video or audio fileor click to choose. JPG, PNG, WebP, HEIC, AVIF, MP4, MOV, WebM, MP3, WAV, M4A and more.

Runs locally. Your file never leaves this device.

What is SynthID?

SynthID is a watermarking technology from Google DeepMind. It hides an imperceptible signal inside AI-generated content: in the pixels of images and video frames, and in the waveform of audio. Unlike a visible logo or file metadata, the mark travels with the content itself, so it can survive compression, resizing, cropping and filters.

A SynthID Detector looks for that signal. Google opened its official SynthID Detector to everyone on 7 October 2026 at synthid.com, and the same check is built into the Gemini app. OpenAI, NVIDIA and Kakao now watermark their output with SynthID too.

Three signals, three strengths

SynthID watermark: Inside the pixels or audio. Survives most edits. Only Google's detector can read it.
Content Credentials (C2PA): Signed metadata that names the tool and the company. Precise, but lost on screenshots and most social uploads. Read on this page.
AI labels (IPTC and XMP): Plain metadata fields such as "Made with Google AI". Easy to read and easy to remove. Read on this page.

How to use the SynthID Detector

1
Read the metadata here
Drop the original file into the checker above. It lists any Content Credentials, the company that signed them, and AI labels such as Made with Google AI. Nothing is uploaded.
2
Open Google's SynthID Detector
Go to synthid.com and sign in with a Google, Apple or OpenAI account. The service is free, works worldwide in English, and has a daily check limit.
3
Upload the best copy you have
Use the original download rather than a screenshot or a re-shared copy. Heavy compression and editing weaken the watermark, and Google asks for the highest-quality file available.
4
Read the result
Images get a detected or not detected answer. Video and audio results mark the segments that carry the watermark. Compare the answer with the metadata report before you decide.

What the SynthID Detector can check

Content	Google SynthID Detector	Gemini app	This page
Image	JPG, PNG, WebP, HEIC, AVIF, TIFF, GIF, BMP. Up to 100 MB.	Google AI images	Content Credentials, IPTC and XMP labels, generator tags
Video	MP4, MOV, WebM. Up to 100 MB. Marks watermarked segments.	Google AI video under 90 seconds	Content Credentials (for example from Veo), container tags
Audio	MP3, WAV, OGG, FLAC, AAC, M4A. Up to 100 MB. Marks segments.	Google AI audio under 1 hour	Content Credentials (for example from Lyria), metadata tags
Text	Not supported	Not supported	Not supported

Limits as published by Google in October 2026. The Gemini app allows about 10 checks of each type per 24 hours.

AI models that embed SynthID

Content from these sources should register with the SynthID Detector, unless heavy edits removed the mark.

Google: Gemini image models (Nano Banana, Nano Banana Pro, Nano Banana 2), Imagen, Veo video, Lyria music, NotebookLM Audio Overviews and Reimagine edits in Google Photos.
OpenAI: Images from ChatGPT, Codex and the OpenAI API, which also carry Content Credentials.
NVIDIA: Video from the Cosmos world foundation models.
Kakao: Content from Kakao's generative AI services.

How to read a SynthID Detector result

Watermark detected

A SynthID mark is present. The content was likely generated or edited by a SynthID-enabled model. The result does not say how much was changed.

Watermark not detected

No confident signal. The file may be human-made, made by a model without SynthID, or edited heavily enough to weaken the mark. It is not proof of authenticity.

Segment results

For video and audio the detector marks the time ranges that carry the watermark. A short marked stretch inside a longer recording suggests AI material was mixed in.

Rare false positives

Google notes that in very rare cases SynthID can trigger on content without a watermark. Gather more evidence before acting on one result.

SynthID Detector vs. AI content detectors

Question	SynthID Detector	General AI detectors
How it decides	Finds a signal embedded when the content was made	Guesses from visual or statistical patterns
What it covers	Models from Google and the partners that adopted SynthID	Any generator, in principle
Reliability	High for covered models, robust to common edits	Varies widely, with false positives on edited photos
A negative result means	No covered watermark was found, nothing more	A lower probability score