Cut out backgrounds from GIFs and video, frame by frame

A still cutout only has to be right once. An animated one has to be right consistently, and inconsistency reads as flicker — an edge that shifts by two pixels between frames is barely visible in a still and extremely visible in motion.

That is why a per-frame still remover applied naively to video looks bad even when every individual frame looks fine. Temporal stability is a separate problem from edge quality, and it needs a model that carries state between frames.

The GIF background remover processes each frame and exports either a transparent GIF or an APNG. Reach for APNG whenever you can: GIF transparency is binary — a pixel is either fully opaque or fully transparent — so soft edges become jagged ones, while APNG carries a real alpha channel.

The trade is support. APNG is widely but not universally supported, so check the destination before committing. Where GIF is required, expect harder edges and choose a subject with a clean silhouette.

The video background remover is tuned for people matting, and the engine is recurrent — it uses information from previous frames to keep the matte stable. That is what produces green-screen results without a green screen.

Being tuned for people is a real constraint rather than marketing. A clip whose subject is a product on a turntable or an animal is outside what the model was built for, and results will be less reliable than they are on a person.

Cutout quality is decided as much by the source as by the model. Even, diffuse lighting on the subject, meaningful separation between subject and background, and a background that is not the same tone as the subject all help more than any setting.

Motion blur is the usual killer. A fast pan or a quick gesture produces frames where the subject boundary genuinely is ambiguous, and no matting model can recover an edge that was not captured.

Once a clip has a transparent background, the video editor at /video-editor places scene backgrounds behind it, and can do the removal in-editor as well. That is usually the faster path for anything you are going to composite anyway rather than exporting and reimporting.

The video object remover is the related tool for the opposite job: click an object in a clip, have it outlined, and remove, blur, or pixelate it with tracking across frames.

Transparency in video means WebM — H.264 MP4 has no alpha channel, so exporting a cutout to MP4 flattens it against a background. If you need to keep the transparency, WebM is the format.

If the destination demands MP4, composite the background you want first and export the finished frame rather than expecting to composite later. And as always, everything here runs on your device: the clip is not uploaded.