Change the background of your video.

Upload the clip, write where it should be instead, and say what has to stay. The red-canyon edit on this page started from two sentences: “Change the background to a red sandstone canyon at golden hour. Keep the walking woman, her clothes, her full-body framing and the original camera movement.” Aisma shows the price in credits before you generate, and the finished video lands in your private Library.

It shouldn’t, because nothing is cut out. Background-remover apps lift the person off the frame and set them over a new picture, which is where hard edges and mismatched light usually come from. Aisma renders the new place and the person together in every frame. In the walk on this page, her beige sweater picks up a blue-green tint in the underwater tunnel and a warmer tone in the red canyon.

Yes. In the corridor walk the camera moves back ahead of her for the full 10 seconds as she walks toward it, and the tunnel, the forest and the canyon slide past at the same pace. The car edit has a slow move along the fender. Put “the original camera movement” on your keep list, as every brief on this page does.

Yes. Add a photo of the spot (up to 10 photos per request; JPEG, PNG or WebP, 20 MB each at most) and point to it by number: “use the beach in Photo 1 as the new background.” Aisma treats the photo as a guide for the scene rather than pasting it in, so expect your place rebuilt around the person, with some details that differ. None of the four backgrounds on this page used a photo; each started from a written request.

Aisma hands back a finished video with the new place already in it, so there’s no keying or compositing step afterward. If you were going to drop the person onto a scene of your own, describe that scene in the request and let Aisma build it around them. Transparent and green-screen exports are on our roadmap. If your work depends on one, tell us what you’d make with it at support@aisma.ai.

Clips where the person stands clear of the background and the camera moves steadily. Our corridor walk is a good model: one person seen head to toe, with the camera tracking at her height. If the framing matters, say so; the canyon brief asked to keep “her full-body framing,” and the result did. Before you post, check faces and any lettering on signs, since generated details can vary. If one object is the problem rather than the whole place, remove it instead.

Clips of 2–60 seconds, up to 500 MB, in MP4 or MOV, vertical, square or horizontal. You get an MP4 at 720p or 1080p with the same length and framing as your clip and its original sound. For a longer video, change the background in parts of up to 60 seconds.

Aisma is paid with credits, and the price follows your clip’s length and the quality you pick, 720p or 1080p. The editor shows the exact number before you generate. Each shot between cuts is priced in whole seconds with a 2-second minimum, so a fast-cut montage costs a little more than its length suggests, and a part that fails costs nothing. On the web, a new account can prepare its first three edits before paying: Aisma studies the clip and asks a few quick questions. Rendering uses credits. Current rates are on our pricing page.

Edits take minutes, and longer clips take longer. The task page shows progress. On the web, Aisma may ask a quick question before it renders and goes with its best reading if you don’t answer within about five minutes. The finished video waits in your private Library. Without an active weekly plan it stays there for 72 hours, so save it to your device.

Yes. Aisma runs in the iPhone app (iOS 16.4 or later), in the Android app and in the browser at aisma.ai. Film the clip, open it in Aisma, type the new place and generate. Once you sign in, your Library and credits follow you to every device.

Yes. The corridor walk in all three worlds and the car on the parking deck are real Aisma edits of the clips shown with them, and Full Aisma edit plays each one whole. Examples marked Previous version were recorded with a previous version of Aisma. Before → After views show the original and the edit at the same moment of the clip, never cropped, sped up or retouched. Your result depends on your clip and request.