Loading tool…
You don’t need a green screen to change what’s behind you on camera. Drop in a clip, watch the new background appear live next to the original, and export an MP4 or a transparent WebM. Every frame is processed on your own device.
Processing locally on your device
Overview
Removing a background from video means running person segmentation on every frame, usually 24 to 60 times per second, and cutting the person out with the result. This tool uses Google’s MediaPipe Image Segmenter with one of two small models. The Fast option, MediaPipe Selfie Segmenter, is about 0.25 MB and is made for people facing a webcam or phone. The Quality option, Selfie Multiclass, is about 16 MB and separates hair, skin and clothing from the background, which gives cleaner edges around hair and loose clothes but runs roughly two to four times slower. Both are Apache-2.0 licensed. They download once along with the MediaPipe runtime (about 12 MB) and are cached after that.
Each frame is shrunk to 320 pixels on its long side for the model, which works at 256 × 256 internally. The confidence mask it returns goes through an adjustable threshold ramp, gets scaled up to the output size with smoothing and an optional feather, and then cuts the person out of the full-resolution frame. Temporal smoothing blends each mask with the previous one to reduce the edge shimmer you get from per-frame segmentation. The model runs on your graphics card through MediaPipe’s WebGL delegate when it can, and falls back to WebAssembly on the CPU when it can’t.
For export, the tool seeks a hidden copy of the video frame by frame and composites each frame on a canvas. FFmpeg, compiled to WebAssembly, then encodes the frames and copies in the original audio. The frame rate follows the source up to 30 fps. Output is capped at 480p, 720p or 1080p (on the short side) and is never upscaled. Expect roughly 5–20 frames per second for the compositing pass on a typical laptop, followed by the encode, so a 30-second 720p clip usually takes a few minutes. The tool shows the frame count and an estimate of the memory needed before you start, and it warns you about long ranges.
You have five background choices. Blur keeps the room recognisable but soft. Colour gives a flat studio backdrop. Image takes any photo, scaled to fill the frame. Green screen fills the background with chroma-key green (#00B140), so an editor can key it out later. Transparent exports a WebM (VP8) with an alpha channel. That file plays in Chrome, Edge and Firefox and can be layered in editors that read WebM alpha. It is several times larger and slower to produce than MP4, because each frame is kept as a lossless PNG until it is encoded.
| Background | Output | Best for | Notes |
|---|---|---|---|
| Blur | MP4 | Talking heads, keeping context | Hides small mask errors best |
| Colour | MP4 | Clean studio look, brand colours | Edges are most visible on high-contrast colours |
| Image | MP4 | Virtual sets, campaign visuals | Image is scaled to cover the frame |
| Green screen | MP4 | Keying later in an editor | #00B140 chroma green |
| Transparent | WebM (VP8 alpha) | Overlays and compositing | Larger files, slower export |
Drop in a video, ideally with one or two people clearly in frame.
Wait for the AI model to load, then play or scrub the original. The preview on the right updates live.
Choose a background (blur, colour, image, green screen or transparent) and adjust the mask, feather and temporal smoothing.
Drag the handles on the timeline to export just the part you need, and pick 480p, 720p or 1080p.
Click Export, keep the tab open while the frames are processed and encoded, then preview and download.
Use cases
Swap a messy home office for a clean colour or a soft blur before you publish a course video or a product walkthrough. You don’t need lights or a backdrop.
Export the presenter on chroma-key green or as a transparent WebM, then drop the clip over slides, gameplay or B-roll in your editor.
Hide family photos, whiteboards or anything else behind you before sending a recorded message to a client or posting a clip publicly.
Put a speaker in front of a brand-colour background or a campaign image so short clips from different shoots look consistent.
Pro tips
Answers to common questions about the video bg remover.
Compositing runs at roughly 5–20 frames per second, depending on your device, the model and the resolution, and encoding adds more time on top. A 30-second 720p clip at 30 fps has 900 frames, so plan on a few minutes. A 1080p export or the Quality model takes longer. The status panel shows the live frame rate.
Yes. Both models segment every person in the frame, not just one. Results are best with people at a typical webcam or interview distance. Very small figures in wide shots are often missed.
The models work at 256 × 256 pixels internally, so the mask is scaled up to your video’s resolution. Use the Feather and Mask sliders to tune the edge. For fine hair, the Quality model does noticeably better than Fast.
Yes. Choose Transparent and the export is a WebM with an alpha channel (VP8). It plays with transparency in Chrome, Edge and Firefox and imports into editors that support WebM alpha. If yours doesn’t, export Green screen and key it there.
Yes. The original audio track for the selected range is copied into the result: AAC for MP4, Opus for WebM. Videos without audio export without it.
No. ToolEnhancer processes files on your own device inside the browser. They are never sent to, stored on or seen by our servers.
Yes. The video bg remover is free to use with no sign-up, no watermarks and no daily limits.