TonySkill

Remove an object from a video

Remove an object, a person or a logo from a video with one click: the AI tracks it and replaces it with the real background found in the video. Sound kept, free, no watermark.

Tool in improvement

Leave your email and we'll let you know as soon as it's ready.

Your address is only used to tell you when this tool is back.

How to remove an object from a video

  1. Drop your video (30 seconds max; result in 720p).

  2. Pause it on a frame where the object is clearly visible and click it.

  3. Click “Erase the object”, then download the video once processing is done.

Why erase here

One click
Click the object or person: the AI (SAM 2) tracks it from frame to frame, even when it moves.
The real background, not a guess
The hidden background is first found in the video's other frames; the AI only fills what was never visible.
Sound kept, sharp 720p result
The soundtrack is kept (converted to AAC so it plays everywhere); the result is an H.264 MP4, 720 px on the short side, up to 30 frames per second.
Nothing is kept
The video is deleted once processed, the result after you download it.

To remove an object or a person from a video for free, drop an MP4, MOV, M4V, WebM or MKV of up to 30 seconds and 95 MB above, pause it on a frame where the object is clearly visible, tap it (up to 8 points on the same object), then click “Erase the object”: an AI tracks it through the video and fills the area with the background found in the other frames. You get a 720p MP4 with the sound kept, no account and no watermark added by us, but the free use is limited to 2 videos a day per browser and can be refused when the site's daily budget is used up.

What the tool can and cannot do

The eraser works in two steps. First, an AI model (SAM 2) follows the object you tapped from frame to frame, even when it moves. Then each masked area is filled with the real background, taken from other frames where that part of the scene was visible; only what is never visible anywhere in the clip is invented by a second AI (MI-GAN). That is why results vary a lot from one video to another.

Typical caseWhat to expect
A person walking through the frame, or a moving cameraUsually the cleanest case: the background behind them is seen in other frames and is reused as is
A small object that moves (a ball, a bird, a passer-by in the distance)Generally good, as long as you tap it on a frame where it is sharp and clearly separated from the background
An object that never moves, filmed by a fixed camera (a sign, a bin in front of a wall)The background behind it is never seen, so the AI has to invent it: expect a blurry or slightly shaky patch
A very large object covering most of the imageHard: there is little real background to reuse and much to invent
Several objectsOne object per pass: download the result, drop it again and tap the next one

Always watch the whole result before using it: the page itself warns that areas rebuilt by the AI may look blurry or unstable.

Step by step

  • Drop your video: MP4, MOV, M4V, WebM or MKV, 95 MB and 30 seconds at most. If your clip is longer, cut the part you need first with Trim video; if it is too heavy, shrink it with Compress video.
  • Find a good frame: press “Play the video”, then pause on a frame where the object is fully visible, sharp and not hidden by something else.
  • Tap the object: a marker appears. For a large object or a person, tap several parts of it on the same frame (head, body, legs): up to 8 points. “Start the selection over” clears them.
  • Click “Erase the object”: the page shows the upload, then your place in the queue, then the tracking and erasing progress.
  • Download: the result is an MP4 named after your video with “-erased” at the end. “Erase something else” lets you start again on the same video.

The page estimates 1 to 2 minutes of processing for a 30-second video, up to 4 to 5 minutes when the GPU server has to wake up, plus the wait if other videos are ahead of yours. Keep the tab open until the result arrives.

Limits and free quota

  • Duration: 30 seconds maximum. A longer video is refused before upload, with a message pointing to the trim tool.
  • Size: 95 MB maximum.
  • Result: MP4 (H.264) in 720p, 720 pixels on the short side with the original proportions, so a vertical phone video stays vertical; at most 30 frames per second. The sound is kept, re-encoded in AAC.
  • Free tries: 2 videos per day per browser. When they are used, the page offers to leave an email address for 3 more tries the same day (5 in total).
  • Site-wide daily budget: the GPU is rented by the second, so there is a daily spending cap for the whole site. When it is reached, the tool refuses new videos until it resets, even if you have not used your own tries. This happens on busy days.
  • Reset time: quotas restart every day at midnight UTC; the error message gives that time in your own time zone.
  • One video at a time: while one of your videos is being processed, a second one is refused.
  • Failed attempt: if processing fails for a technical reason, or the result has expired, the try is given back so you can run it again.

The tool can also be temporarily switched to maintenance; the page then says so instead of showing the upload area.

Tips for a cleaner result

  • Pick the frame carefully: the tracking starts from the frame you tap. A frame where the object is blurred by motion, cut off by the edge or overlapping another object gives a worse mask.
  • Tap several points on a person: one tap often selects only part of the body; add points on the head, torso and legs.
  • Keep the clip short and focused: only send the seconds where the object appears. Less video also means a shorter wait.
  • Prefer moving shots: when you have the choice between two takes, the one where the camera or the object moves gives the AI more real background to work with.
  • iPhone videos: MOV files from an iPhone are accepted as they are, including vertical ones; the result keeps the vertical format. If the server says it cannot read the file, convert it to MP4 with MOV to MP4 and try again.
  • Text, subtitles or a logo burned into the image: see Remove subtitles from video. Only remove a watermark or logo from videos you own or have the rights to.

Where your video goes and when it is deleted

This processing is too heavy for a browser, so your video is uploaded. It goes to our compute server, which hands it to a GPU computing service (Modal, in the United States) where the tracking and erasing run. The copy of your upload on our compute server is deleted as soon as it has been handed to the GPU. The finished result is brought back to our compute server so you can download it, and erased automatically about an hour after the end of processing, or 30 minutes after your last download if that is later. We add no watermark, and no account is needed; an email address is only asked if you want the 3 extra tries.

Good to know

  • If the page says the server is off, busy or the queue is full, wait a few minutes and try again. When processing fails for a technical reason, your free try is given back.
  • The result is in 720p even if your video was 1080p or 4K: keep the original if you need full resolution elsewhere.
  • To blur faces instead of erasing them, or to remove the whole background, see the related tools below.

Something wrong with this tool?

Frequently asked questions

How do I remove a person from a video?

Drop the video, click the person on a frame where they're clearly visible, then click “Erase the object”. An AI (SAM 2, by Meta, open source) tracks them through the video; each area is then filled with the background found in the other frames, and whatever was never visible is completed by a second AI (MI-GAN).

Is the result perfect?

Not always, and we'd rather say so. When the camera or the object moves, the background behind it is often visible elsewhere in the video and the result is very clean (in our test, 86% of the area was filled with the real background). When an object stands still in front of a background that's never seen, the area is made up and may look blurry or shaky. Paid tools use heavier video models under non-commercial licences.

Why 30 seconds max?

Tracking and erasing run on a GPU rented on demand, one video at a time: about 3 seconds of processing per second of video. Beyond 30 seconds, the wait would get too long for everyone. The result is 720p (1280 px wide at most), 30 frames per second at most.

Can I erase several objects?

One object per pass: download the result, drop it again and click the next object.

Is my video kept?

No. It is sent to a GPU computing service (Modal, in the United States), processed, then deleted right away. The result stays on our server for 60 minutes after it is ready (30 minutes after your last download, so you can download it again), then it is deleted.

Related tools

  • Remove object from photoClick an unwanted object (or brush over it): the AI detects it, erases it and rebuilds the background behind.FreeNo sign-up
  • Remove video backgroundReplace a video's background (blur, colour, image or green screen), no special setup.FreeNo sign-up
  • Blur facesAutomatically blur the faces in a photo or video, and license plates with a simple box.FreeNo sign-up
  • Video editorEdit videos in your browser: trim, join, speed up, 9:16 TikTok format, text, subtitles, music.FreeNo sign-up
  • Trim videoShorten a video to the part you want, to a tenth of a second, or instantly without re-encoding. MP4, MOV, WebM.FreeNo sign-up
  • Compress videoShrink MP4, MOV, WebM or MKV videos by up to 90% with no upload: everything runs in your browser.FreeNo sign-up