Confirm there is a mark
Some Grok Imagine images show no visible wordmark at all. If yours is already clean, there is nothing to erase, and editing would only risk softening the corner.
This Grok watermark remover is a browser editor for painting over visible text on a still image and letting an AI model rebuild what was underneath. Grok Imagine outputs do not all look the same, so check your own file first, then mark only the area you actually need to clean.
JPG, PNG or WebP. Drop, paste or pick a file. It is processed in your browser.


xAI confirms that Grok Imagine adds a watermark, but it does not publish what the mark looks like or where it sits. What is known comes partly from xAI and partly from outside observers, and the two do not fully agree.
Because the visible Grok Imagine watermark is undocumented and appears inconsistent, treat the bottom-right position as a hint, not a rule. Zoom into your own image before you start masking.
Everything happens on this page. The picture is decoded and repaired inside your browser rather than on a server.
Use the Upload button, drag the file in, or paste it straight from the clipboard. JPG, PNG and WebP files are accepted.
Look at the corners for a wordmark. Press "Auto-mark text" to highlight text regions in general, then use Mark and Unmark with the brush to keep only what you want gone.
Press Erase. The inpainting model fills the masked pixels using the texture around them. Undo and redo let you step back if a pass looks off.
Zoom in on the repaired area, touch up any leftover outline, then download the result as a PNG.
Photoreal scenes with sand, sky or foliage behind a small label are where inpainting does its best work, because the model has plenty of similar texture to copy from.


Small choices about the mask and the model matter more than how many times you press Erase.
Some Grok Imagine images show no visible wordmark at all. If yours is already clean, there is nothing to erase, and editing would only risk softening the corner.
Paint a few pixels past the edges of the letters. Anti-aliased text leaves a faint halo that a tight mask can miss.
Quality mode runs the larger model and copes better with fine detail such as sand ripples or hair. Fast mode is handy for a quick preview on slower devices.
No. xAI’s Grok FAQ states that Grok Imagine images and videos include a Grok watermark and that no setting turns it off.
xAI’s Acceptable Use Policy forbids removing the Grok watermark, so erasing it from Grok outputs can breach xAI’s terms. This editor is meant for images you have rights to, for example cleaning other text or labels, and you are responsible for complying with xAI’s terms. Keep AI disclosure wherever it is required and never present AI-generated images as real photos.
There is no Grok-specific detector. "Auto-mark text" highlights text regions of any kind, which may or may not include a corner wordmark. Always review the mask and adjust it with the brush.
This editor does not try to detect or remove invisible provenance marks. Because the result is a new PNG drawn from the browser canvas, the original file’s metadata (xAI’s C2PA credentials and the EXIF signature and UUID) is not copied into the download. That is a side effect of re-encoding, not a way to erase provenance: any watermark embedded in the pixels is not targeted, and xAI keeps its own records of what Grok Imagine generated. Keep the original file and disclose AI generation wherever it is required.
No, it handles still images only. You could edit a single frame or screenshot as an image, but video files are not supported.
Your image never leaves the browser. Only the AI model files download the first time you use the editor, and they are cached afterwards. Masking, erasing and downloading are free with no account.
Each platform marks its images differently. Pick the guide that matches where your picture came from.
Open an image you have rights to, mark the text you want gone and download the repaired PNG, all inside your browser.