A cup in the foreground can compete with the person you photographed. Open Free erase, upload your own image, and paint only the object you want gone. AI fills that area using the surrounding scene. You decide what to remove; this is not a background cutout or a promise to recover hidden original pixels.
One photo at a time · No installation · Check credits before generating
Original
AI editedChoose a JPG, PNG or WebP, or reuse one from recent uploads. Keep a copy of the original.
Paint over an object and its shadow. In this example, the white cup is removed and the table is filled in; the working image outside the painted area is retained.
The floating bar shows the estimate and balance. You need to sign in and have sufficient credits. Once submitted, you can submit another task.
Check the result against the original, inspect edges and faces, and download the version you want. Saved tasks are available in My Works.
Free erase is useful when a distracting foreground prop, loose item or clearly separated background object interrupts a photo you otherwise want to keep. The goal is to reveal a plausible surface, not insert a different prop. If you want a cup to become a plant, choose object replacement instead. If you need a transparent subject for a design, use the separate background remover. These outcomes require different controls and should not be confused.
Start with the least compressed copy you have and keep a separate original. Nearby wood grain, wall colour and surface edges give the fill useful context. A small object against a plain wall is a different job from an object covering a hand, face or complicated lettering. When most of the hidden background is missing, the editor must invent it; a larger file cannot restore information that was never photographed.
Use the brush to cover the whole object, including a handle, thin supports and disconnected parts. Add the object's shadow if you want that shadow removed too. Keep the selection off the person's hand, notebook lettering and laptop edge unless those areas are meant to change. A small margin can prevent fragments, but excessive painting gives the model too much room to invent. Correct accidental strokes before generating instead of relying on the prompt to rescue a careless selection.
For a crowded desk, decide which object matters most and work from the original rather than repeatedly editing a heavily altered export. Several overlapping objects can share shadows and obscure each other. Removing all of them in one pass may leave an implausibly smooth patch. Separate attempts help you see which selection caused the problem; they are separate generation tasks, not a batch of photos included in one request.
Look first for a remaining cup handle, a duplicate edge or a shadow with no object attached. Then follow the notebook border and the direction of the desk grain across the filled area. At normal viewing size, ask whether the space looks naturally empty. At a closer view, look for repeated texture or an abrupt change in sharpness. For a visibly broken straight line, repaint a tighter region from the original; do not hide the fault behind a smaller preview.
Brush-based erase retains pixels outside the marked area of the working image. That working image can be resized for processing, so this is not a claim that every pixel in your full-resolution camera file stays identical. The boundary is blended, and pixels you painted are intentionally replaced. Whole-photo cleanup actions and text instructions use a different scope and do not offer the same outside-selection protection. Choose Free erase for a specific target beside details that matter.
Do not erase features that make a listing, record or personal photograph misleading. Editing a visible defect out of an item for sale can change what a buyer understands they are getting. Keep documentary originals and identify substantive alterations when appropriate. Use photos you own or have permission to edit. A removed object can still leave a reflection or partially occluded surface that needs inspection; the result is a creative edit, not factual evidence of the scene.
Compare the original and result before adding an art filter or upscaling. A painted effect can conceal a rough join without fixing it, and enlargement can make a bad fill more obvious. Keep the original and your chosen edited version separately. Download availability follows your account plan. If the task is rejected or fails, follow the displayed status rather than assuming a usable result exists; don't repeatedly submit the same request without reading the message.
Generation requires login and sufficient credits. Check the estimated cost before each submission; extra attempts are separate tasks. Download access follows your plan. Uploaded material is processed by the generation service. Saved tasks can be deleted from your account; see the privacy policy for retention and deletion details.
Privacy and deletion ↗Not for Free erase. Upload a photo, brush the target and submit after checking the cost. A brush selection tells the editor where to remove content. Do not switch to text instructions just to avoid covering a difficult outline.
No. The fill is generated from visible context. An object covering a specific word, facial detail or unique pattern can hide information that cannot be recovered reliably. Keep the original if factual accuracy matters.
You can mark an unwanted background element, including a clearly separated person or date. Protect the main subject and use only authorised material. Tiny marks, crowded groups and overlapping faces require particular care; no universal success rate is promised.
One photo at a time · No installation · Check credits before generating