AI Magic Eraser 2026: Remove Any Object from Photos and Videos (Free, No Photoshop)

The term got popular thanks to Google Photos' Magic Eraser and Samsung's Object Eraser on recent Galaxy phones. Under the hood, they all do the same thing: you indicate a region of a photo (a person, a sign, a wire), and a generative inpainting model regenerates the pixels underneath so the region looks like it was never there. Photoshop ships the same concept as Generative Fill.

Genso AI's Magic Eraser — https://gensoai.io/eraser — brings that workflow to any browser, for free, and extends it to video with temporal inpainting.

The practical list is long:

In photos — strangers in the background, photobombers, ex-partners from trip photos, tourists in front of landmarks, wires and cables, reflection artifacts, date stamps, street signs, trash, props, accidental brand names, old text overlays, logos, and watermarks.

In videos — the same list, plus moving objects like cars, people walking through a shot, boom mic tips, crew reflections, and set clutter. Because the video model is temporally aware, the mark stays gone across every frame without popping or shimmering.

The one category where you should use a different tool is platform watermarks (TikTok, Shutterstock). Those have dedicated detectors in the Watermark Remover that beat a generic eraser.

Step 1 — Open /eraser, pick the Image Eraser model.

Step 2 — Upload the photo (JPG / PNG / WebP, up to 4K).

Step 3 — Click "Paint mask on image". The mask editor opens with your photo loaded.

Step 4 — Brush over the object. Resize the brush for precision on small things or large strokes on big ones. Undo is always one click away.

Step 5 — (Optional) Type a prompt describing what should replace the region ("a plain sand beach", "clean brick wall"). You can also skip the prompt entirely.

Step 6 — Generate. The AI inpaints the masked region with matching lighting, texture, and perspective.

Step 7 — Download the cleaned photo in the same resolution as the original.

Step 1 — Open /eraser, pick the Video Eraser model.

Step 2 — Upload the clip (MP4 / MOV / WebM, up to 10 minutes).

Step 3 — The first frame is used as the mask canvas. Paint over the object you want gone on that frame — the mask will be applied temporally to every subsequent frame.

Step 4 — (Optional) Add a prompt describing what to remove ("the person in red") — useful when the object moves a lot, because the prompt steers the model at frames where the fixed mask is slightly misaligned.

Step 5 — Generate. Processing time scales with clip duration.

Step 6 — Download the cleaned MP4. Because the underlying model uses temporal inpainting, the removed region stays gone without flicker or ghosting.

Tip: the video eraser works best for static or slow-moving objects (logos, signs, stationary people in the background). Very fast-moving objects benefit from a tighter mask and an additional pass if needed.

vs Cleanup.pictures — Cleanup.pictures kicked off the free-magic-eraser wave. Great quality on images, but free-tier is low-resolution and there's no video. Genso AI's free tier keeps full resolution and adds video.

vs LaMa Cleaner — LaMa is excellent if you want to self-host. Genso AI trades self-hosting for zero setup, full-resolution output, and a matching video eraser inside the same studio.

vs Photoshop Generative Fill — Photoshop is the gold standard for professional retouchers but costs a subscription and has no video eraser in the same workflow. Genso AI is "Generative Fill in the browser" — good enough for 90% of practical edits, free.

vs Google Photos Magic Editor — Magic Editor is excellent on a Pixel phone but confined to Google's ecosystem. Genso AI runs on any device with a browser and extends to video.

vs Samsung Object Eraser — Same "on-device only" limitation as Magic Editor. Genso AI is device-agnostic and video-capable.

Mask only — Use when the object is static and you can outline it precisely (e.g. a trash can in a landscape photo, a logo in the corner of a product shot).

Prompt only — Use when the object is hard to paint (a moving person in a video, a thing made of many small parts) but easy to describe ("remove the dog", "remove the person in red").

Mask + prompt — The best results on tricky scenes. The mask tells the AI *where*, the prompt tells it *what*. Classic example: a dim scene with overlapping people; mask the one you want gone, prompt with "the person in the blue jacket in the center" — the AI locks onto the correct subject instead of guessing.

Travel photos — Remove tourists from the Colosseum shot. Takes 20 seconds.

Real estate listings — Erase clutter, cables, and personal items for cleaner hero images. A whole set of listing photos can be cleaned in 5–10 minutes.

Product photography — Remove dust, stray hairs, or studio reflections without manual cloning.

Video podcasts — Remove a mic stand that snuck into the frame, or clean up a busy background before cutting to social.

UGC ads — Remove competing brand logos on props before running a paid campaign.

Family photos — Erase an ex from a treasured vacation photo. (Yes, this is the #1 actual search query for "remove person from photo AI".)

The Magic Eraser is strongest when chained with Genso AI's other free utilities:

1. Erase clutter from a product photo at /eraser. 2. Remove the background at /background-remover to get a marketplace-ready transparent PNG. 3. Upscale the final asset to 4K at /image-upscale for print or retina displays.

All three tools share the same free credits and the same studio. No installs, no plugins, no Photoshop.