Last updated: July 2026
Take the background out of a video right in your browser — nothing to install. Upload a short clip of a person and download a transparent WebM, a green-screen MP4 or an alpha matte. Free, no sign-up, no watermark.
Remove my video background now →Most background removers want a subscription, stamp a watermark on the export, or only work if you already own a green screen. VideoBGNinja runs Robust Video Matting on our own GPU, so there's no cloud bill per clip and nothing to pay: it separates a person from any background frame by frame, free, with no account and no watermark.
Because it runs online, there's no desktop app, plugin or account: open the page, drop your clip in, and the processing happens on our GPU, not your laptop. That means it's the same whether you're on Windows, macOS, Linux, a Chromebook or a phone.
Every job hands you three files so you're never stuck: a transparent WebM with a real VP9 alpha channel (drops onto any background in browsers, OBS and modern editors), a green-screen MP4 you can chroma-key in any app, and a black-and-white alpha matte MP4 for precise compositing. Your original audio stays in the WebM and the MP4.
For a website or an OBS overlay, reach for the WebM; for a video editor that can't read alpha, use the green-screen MP4 and key it out. The OBS overlay guide shows the streaming setup end to end.
The tool is tuned for short clips: up to 1 GB and 60 seconds per upload, processed at up to 1280px. Longer material? Trim it into parts first — that keeps processing fast and free for everyone. It's trained on people, so a human subject is what it does best.
Nothing here is hidden behind a sign-up, so it's worth knowing the ceilings before you upload. A clip can be up to 60 seconds and up to 1 GB. Sixty seconds sounds short until you notice how video overlays actually get used — an intro, an outro, a lower third, a reaction clip, a product spin. Those are all well under a minute. If you have a five-minute recording, cut out the ten seconds you need first; every editor and even your phone's photo app can trim without re-encoding, and trimming before upload is faster than uploading the whole thing.
The longest side of your video is scaled down to 1280 pixels if it's bigger. That sounds like a loss and usually isn't, because a cut-out person almost never fills the frame at full size — they sit in a corner of a slide, over gameplay, in a picture-in-picture box. At that size 1280 is more resolution than the box you're dropping it into. It also keeps matting fast, which is why a minute of video comes back in around a minute instead of twenty.
Container-wise it's relaxed: MP4, MOV, WebM, MKV, AVI, M4V, MPG, MPEG, WMV and 3GP all go in. Phone footage shot in portrait comes out portrait — the rotation flag that trips up so many tools is read and applied, so you won't get your video lying on its side. Audio, if the clip has any, is kept in both the green-screen MP4 (AAC) and the transparent WebM (Opus). And there's one active job per visitor at a time: video matting is heavy, and a queue that lets one person start eight uploads makes everyone else wait. Finish one, start the next.
The traditional way to remove a background without a green screen is rotoscoping: draw a mask around the subject, then adjust that mask over and over as they move. After Effects has a Roto Brush that helps, DaVinci has a magic mask, and both are genuinely good — but both are you, sitting there, correcting frames. For sixty seconds at 30fps that's 1,800 frames to be responsible for.
What runs here instead is Robust Video Matting, a model built specifically for pulling a person out of video. The important word is video: it doesn't treat your clip as 1,800 unrelated photos. It carries state from one frame to the next, so the edge it found around your shoulder in frame 200 informs frame 201. That's why the result doesn't flicker the way frame-by-frame photo cut-outs do, where the outline wobbles a pixel or two every frame and the eye reads it instantly as fake.
Where a manual roto still wins: a subject that isn't a person, two people overlapping and swapping depth, an object passing in front of the subject, or a shot where you need a mask on something specific like just a hand. This model knows people extremely well and everything else only roughly. If your clip is a person talking, moving, dancing or presenting, the automatic result is usually better than a rushed manual mask — and it's done before you'd have finished the first pass.
No. It runs entirely in your browser — upload a clip and download the result. Nothing to install and no account.
Yes. Any modern browser works, since the heavy lifting happens on our GPU, not your device.
Up to 60 seconds and 1 GB per clip. Trim longer footage into parts first.
No. It's deleted right after processing and results are wiped within 45 minutes.