AI Background Fill
The editor uses surrounding colors, textures, and shapes to generate a replacement for the selected area.
FREE AI PERSON REMOVER
Remove a passerby or other unwanted person from a photo online free. Mark the person with the brush and let EraseKit fill the selected area using nearby image context, right in your browser.
SIMPLE, PRIVATE PHOTO CLEANUP
Mark the area you want changed, let AI inpainting create a fill, and decide whether the result suits your photo.
The editor uses surrounding colors, textures, and shapes to generate a replacement for the selected area.
Image processing runs locally in your browser, without sending the photo to an EraseKit processing server.
Start editing without an account, subscription, or export watermark.
Inspect the edited area and surrounding edges before you decide to keep the result.
STEP-BY-STEP GUIDE
Use a focused selection and check the reconstruction at full size.
Load a PNG, JPG/JPEG, or WEBP image into the editor. Keep your original file as a backup.
Use the brush to cover the full person, including stray edges and any shadow you want removed. Avoid nearby details.
Choose Erase object and give the browser a moment to generate the fill. Large images may take longer.
Check textures, lines, and edges around the repaired area. Download only if the reconstruction works for your use.
BEFORE AND AFTER · ILLUSTRATIVE EXAMPLE
In a simple scene, the marked person-sized area is replaced with a generated continuation of the visible background. The hidden scene is not restored from original data; results depend on what remains visible around the selection.
Try the editorA passerby interrupts the view across a simple background.
The selected area is filled from nearby visual context.
BETTER SELECTIONS, BETTER CONTEXT
Removal tends to be more convincing when the person occupies a modest area and the nearby scene gives the model texture or color to continue, such as sky, a wall, grass, or soft water.
Include the whole outline and contact shadow, but keep the mask close enough to avoid unnecessarily replacing nearby signs, faces, or architectural detail.
A person covering a substantial part of the frame, crossing a railing, or standing in front of patterned tiles leaves little context to guide a plausible fill. Hair, thin branches, and sharp edges can also be difficult.
Try a revised selection if a fragment remains, then inspect again. Each attempt is a new estimate; repeated runs do not reveal the hidden original.
PRACTICAL SELECTION TIPS
Start with the person’s full outline, then keep the selection tight enough to avoid replacing background details you want to retain.
UNDERSTANDING THE RESULT
No—the original pixels behind the person are not available to the editor. An image inpainting model generates a plausible fill from visible surroundings, so a result can look reasonable without matching what was really there. Research on image inpainting and masked-region completion describes the broader technique; it does not make a reconstructed image a record of hidden details.
For a casual photo, you can decide whether the visual result is useful. For documentary, news, archival, or evidentiary images, keep the original and do not present generated content as untouched.
MORE PHOTO CLEANUP HELP
KNOW THE LIMITS
Use edits responsibly. For documentary, evidentiary, or other contexts where an accurate record matters, do not present reconstructed content as an untouched photograph.
BROWSER-BASED PROCESSING
EraseKit performs the image-editing step in your browser using ONNX Runtime Web. Its official browser documentation explains how machine-learning models can run in web applications. Your browser still needs to load the site and model resources, and device memory or browser support can affect processing.
Browser-based image processing is separate from the website's other data practices. Read the Privacy Policy for details about analytics, browser storage, and network requests.
COMMON QUESTIONS
Open the editor on this page, select a PNG, JPG, or WEBP image, brush over the person and any shadow you want removed, then choose Erase object. Review the repaired area before downloading. EraseKit is free to use without an account.
Not necessarily. The original background behind the person is hidden, so the AI generates a plausible fill from visible surroundings. It cannot recover the exact pixels that were covered.
A smaller person against a relatively consistent background, such as open sky, a wall, or a soft landscape, is usually easier than a person covering detailed architecture, text, faces, or repeating lines.
The editor can apply a selection at the same relative coordinates across a batch, but it does not track a person as they move. Use it only when the subject and framing stay in the same place; otherwise edit each photo separately.
The image editing and inpainting run in your browser. Your browser still loads the website and model resources; see the Privacy Policy for details about site operations. ONNX Runtime Web's browser documentation explains how model inference can run in web applications.
Large selections, complicated backgrounds, and objects crossing strong edges leave less useful context for reconstruction. A tighter selection, a suitable source image, and careful review can help, but cannot guarantee a seamless result.
READY WHEN YOU ARE
Choose a photo, make a careful selection, and review the AI-generated background before saving.
Open the free editor Free · No account required · Browser-based editing