IMAGE CLEANUP
REMOVE UNWANTED MARKS/TEXT
CLEANUP IMAGE
Upload media to remove unwanted marks/text automatically.
JPG · PNG · HEIC
Removes
How Magic Image Cleanup Works
FauxLens Magic Cleanup uses AI-powered inpainting to remove unwanted elements from images (objects, text overlays, timestamps, channel logos and other visual artifacts) and reconstruct the background behind them. It reads the surrounding pixels to work out what was underneath, so what you get back is a clean image rather than a patch.
Object & Text Removal
Removes objects, timestamps, channel logos, subtitles, and floating text overlays. The inpainting model analyzes the surrounding pixels and reconstructs what the background likely looked like before the overlay was added - producing a clean result that matches the surrounding texture and lighting.
Artifact & Noise Cleanup
Smooths out JPEG compression artifacts, banding, chromatic aberration, and visual noise from low-quality scans or heavily compressed downloads - without destroying important edge detail or fine texture in the image.
Privacy Redaction
Remove faces, license plates, sensitive information, or identifying background details from images before sharing. Unlike blurring tools, inpainting removes the element entirely and fills the gap naturally, making the edit invisible rather than obvious.
Background Reconstruction
When an unwanted person, object, or element occludes part of the background, our model reconstructs the missing scene behind it - restoring street scenes, landscapes, or interior backgrounds to a clean state that looks entirely natural.
How Inpainting Works
AI inpainting works by masking the region to be removed and using a neural network trained on millions of images to predict what the pixels underneath would look like, given the surrounding context. Modern inpainting models use diffusion-based reconstruction - the same core technology behind image generation, applied in reverse. The model considers texture, lighting direction, perspective, and color gradients to produce a fill that is visually consistent with the rest of the image. Upload your image, select the area to remove with our brush tool, and the model handles the reconstruction automatically.
Ethics Policy: This tool is intended for legitimate restoration, privacy protection, and personal use. Do not use it to remove copyright notices, watermarks or attribution from content you do not own, or for any deceptive purpose. Accounts that do are terminated.
Secure & Ephemeral
All processing happens in a secure, isolated environment. Your image is never added to any database and never backed up. It is deleted immediately once the analysis finishes. The result of the analysis is kept separately — see our Privacy Policy.
High-Resolution Support
Works on high-resolution images without downscaling. Upload files up to 30MB and receive a full-resolution cleaned image ready for download.
Precision Brush Tool
Use the in-browser canvas editor to paint a precise mask over only the area you want removed. Adjustable brush size gives you pixel-level control over what gets cleaned.
Frequently Asked Questions
Using the cleanup tool
How to clean up an image
- Drop in a JPG, PNG or WebP.
- Paint over the thing you want removed. Rough is fine; you do not need a careful outline.
- Run it. Most images finish in a few seconds. If the edge of the repair looks off, brush a little wider and go again.
- Download at full resolution.
What it handles well
Anything sitting on a surface the model can continue. Objects over grass, sky, fabric, concrete or water come back almost invisibly, because there is a pattern to extend. Text overlays and channel logos are the easiest case of all.
Where it struggles is structure. An overlay across someone's face, a logo sitting on a line of text, a person overlapping a building's windows: the model has to invent geometry rather than continue it, and invented geometry is where you notice the repair. Mask tightly in those cases and expect to make two passes.
Cleanup and detection
Worth saying plainly, because it catches people out: cleaning an image is a generative edit. If you later run that file through an AI image detector it will flag the cleaned region, and it is right to. Verify first, clean second. If what you actually want is to know whether a photo has been edited by someone else, use the Photoshop detector instead.
Common questions
Yes, to start. A removal costs 20 credits and every device gets 40 free credits, so the first two are free. Signing up does not add more - it moves those credits onto your account. There is no watermark on the output, and you get the cleaned image back at full resolution.
The cleaned region is regenerated, so those pixels are new. Everything you did not mask is passed through untouched at the original resolution. We do not re-encode the whole image to a lower quality, which is what most free online cleaners do.
Yes. The masking brush is built for touch as well as mouse, and the tool runs in the browser, so there is nothing to install on either.
Small and medium objects come back cleanly: a timestamp, a bin, a passer-by, a power line, a stray sign. Large removals depend almost entirely on what is behind them. A person standing against sky or grass is easy. A person standing in front of a face, a sign or detailed architecture is not, because the model has to invent structure it has never seen.
No. Cleanup runs on the same zero-retention path as detection. The file is kept only for the length of the job and deleted when it finishes. Your image is never added to any database, and nothing is used for training.
It changes the answer, and it should. Inpainting is itself a generative edit, so a forensic scan of a cleaned image will correctly report AI-generated content in the region you cleaned. If you need to verify a photograph, verify it before you clean it, not after.
JPG, PNG and WebP. Transparency in a PNG is preserved. Very large files are downscaled for the preview but processed at their original size.