Technical explanations, with no marketing promises, for every family of AI watermark: what it is, where it hides, what genuinely removes it and what does not work.
The six families of AI watermarks (invisible Unicode, C2PA, EXIF/XMP, video metadata, statistical text marks, SynthID), what is genuinely removable and what is not.
Read the guide →SynthID marks images from Gemini, Imagen and Nano Banana. Why cropping and compression are not enough, and what a local tool can genuinely do.
Read the guide →Zero-width spaces, joiners, variation selectors: how text copied from an AI carries invisible characters, and how to clean them without breaking your emoji.
Read the guide →A C2PA manifest records the model, the software and the edit history inside your image. Where it hides in a PNG or JPEG and how to remove it locally.
Read the guide →GPS, camera, software, prompt, author: what your files reveal without your knowledge, and how to clean images, PDFs, DOCX and videos locally.
Read the guide →Sampling watermarks mark token choices, not characters. How the green list works, what rewriting can reduce and what it can never guarantee.
Read the guide →Detection and cleaning inside your browser: no upload, no account, no quota. Your files never leave your device.
Open the studio