Image inspector
The image is read in your browser and never uploaded

Inspect any image: who made it, and what it reveals about you

Drop an image to read its provenance, identify the tool that generated it, extract the prompt if one is stored, and see the location, device and identity it carries about you.

Drop an image here, or click to choose PNG · JPEG · WebP — read entirely on your device

Drop an image here, or click to choose

The team behind this tool can build what you need

If you are working on a website, an academic paper, or a system for your organisation — the Erticaz team offers a free initial consultation.

Chat on WhatsApp

What the image carries about its origin

AI-generated images often carry an explicit declaration of their source. The most widely adopted standard is C2PA (Content Credentials): a cryptographic signature recording who produced the image and with which tool, adopted by OpenAI, Adobe, Google and others. It travels alongside a DigitalSourceType field that states the content was made by a trained model.

Open-source generators go further: Stable Diffusion, ComfyUI and Automatic1111 write the full prompt, the model, the seed and the step count as plain text inside the file. The inspector reads and shows them exactly as stored.

And when nothing is there at all, that is information in itself: social platforms strip metadata on upload, and so does anyone hiding a source. The inspector tells you "removed" rather than "clean".

And what it reveals about you

A photo from your phone carries more than you think: coordinates accurate to a few metres, the device model and serial number, the owner name registered on the system, and the capture time to the second. All of it travels with the file when you send or publish it.

The cleaning here is deliberately selective: it removes location, device, identity and timestamps, and keeps three things intact — image orientation (removing it flips the picture), copyright (a legitimate right of the owner), and C2PA provenance.

That separation is intentional: the tool protects your privacy and does not help you hide that an image was AI-generated. Anyone looking for the second will not find it here.

The limits of what this tool can see

Everything read here lives in metadata — a surface layer removed with one click and forged just as easily. A result alone therefore proves nothing.

A deeper layer never appears here: watermarks embedded in the pixels themselves, such as Google SynthID. Those survive cropping, compression, screenshots and metadata removal, and can only be read with tools from the party that produced them.

FAQ

Is my image uploaded to a server?
No. The file bytes are read directly inside your browser and nothing leaves your device. You can disconnect from the internet after the page loads and the tool keeps working.
Which formats are supported?
PNG and JPEG for both inspection and cleaning, WebP for inspection. Other formats are refused explicitly rather than handing back an uncleaned file.
Does the tool remove the AI marker from an image?
No, by design. Cleaning removes what identifies you — location, device, identity — and leaves C2PA provenance intact. Pixel-embedded watermarks such as SynthID cannot be removed by any metadata operation anyway, so any tool promising that is promising what it does not have.
Why does my image show "metadata removed"?
Because the file carries no metadata. This is very common for images passed through WhatsApp, Facebook or Instagram, since those platforms strip it on upload. It does not mean the image is generated or manipulated.
Is the extracted prompt accurate?
Yes when present, because it is not an inference but text written inside the file by the generator itself. Automatic1111 and ComfyUI store it unencrypted, along with the model, seed and step count.