Core Purpose
EraseObject is an online tool focused on image object erasure, mainly helping users remove unwanted elements from photos, such as people, text, wires, clutter, and various distracting objects. Its core feature is that all AI processing is completed locally in the browser, and images are not uploaded to a server, providing privacy protection advantages.
Main Capabilities
- Object erasure: Supports selecting objects to remove by painting with a brush or clicking, and the on-device AI naturally reconstructs the erased area based on surrounding pixels.
- Smart selection: After clicking an object, a local segmentation model automatically outlines its contour, and users can also fine-tune it by adding/subtracting with the brush.
- Dedicated modes: Provides three dedicated finding modes for people, text, and wires, all running locally, making it convenient to quickly locate and remove common targets.
- Multiple selection tools: Supports brush, lasso, and rectangle selection, with sizes calculated according to photo pixels, and zooming does not affect the coverage area.
- Shadow assistance: After smart selection or people mode, attached shadows are indicated in amber, and users can decide with one click whether to include them.
- Structure repair: For structures such as walls, baseboards, shelves, and horizons, the structure is continued first and then the AI fills it in, ensuring straight edges.
- Multiple result attempts: The same selection can generate up to four candidate results, including wider context, mirror processing, or copying nearby textures, for users to choose from.
- Restore and undo: Supports restoring original pixels with the restore brush, with at least twenty steps of history, allowing single-step undo.
- Pixel-level zoom: Supports zoom from 5% to 1600%, provides a navigator at high zoom levels, and the mask always stays aligned.
- Three comparison methods: Hold to view the original image, drag a split-screen divider, or switch comparison.
- Batch cleanup: Can apply the same mask to multiple photos and download them as a package.
How to Use
- Open a photo: Drag and drop, paste, or select an image, supporting formats decodable by the browser such as JPG, PNG, WebP, HEIC, AVIF, BMP, GIF, and TIFF.
- Select what to remove: Paint with the brush, click with smart selection, or use people, text, or wires modes.
- Local erasure: The on-device AI reconstructs the missing area based on surrounding pixels.
- Retouch: Undo, restore, expand the mask, or try other results.
- Download: Export the processed image at the best available resolution.
Privacy and Offline
Images are decoded, edited, and encoded locally on the device throughout the process, with no upload interface. When using the erasure feature for the first time, the editor will clearly inform you of the model to be downloaded and its size; the model is cached by the browser and can be reused for subsequent edits. Once the erasure model cache is complete, you can open it offline, erase, and export. There is no credit system, and repeated erasure is not charged; there is no server queue, and the device processes immediately after you press erase.
Applicable Scenarios
Product photo cleanup (removing stands, hooks, loose threads, dust, and props while keeping product edges clear), real estate photo cleanup, batch object removal (such as studio stands, sensor spots, and homemade overlays), removing text from photos, removing wires, removing blemishes, removing photobombers, and more.
Limitations
The current information does not specify performance metrics such as the maximum supported image resolution or single-processing speed. Some features depend on specific models and require separate downloads; smart selection and people mode each require a separate model download. Offline use requires completing model caching first.
