Mixer AI
AI video editor: replace the background, clothes and style of a finished clip
Two models change a finished clip: Seedance 2.5 in Video editor mode and Gemini Omni in From video mode. You upload the source, optionally a photo of the new item or background, and write what changes and what stays.
More
Motion, camera and timing come from the source video.
Step by step
Prepare the source: for Seedance 2.5, one video of 4–30 seconds in MP4 or MOV; for Gemini Omni 1.2, up to 15 seconds.
Compress a large file to 720p for Seedance. Cut a longer clip into pieces at shot changes.
Pick the model and mode: Seedance 2.5 Video editor or Gemini Omni From video.
Upload the clip and, if needed, a photo of the new outfit, object or background.
Write the prompt: first the one change, then what stays — framing, camera movement, the person's motion and face, timing.
Name each photo by role: the jacket from image 1. If only part of the clip changes, give its time range.
Check the price before you start, then generate.
In the result, look at the beginning, middle and end: did the face change, did extra objects appear, did the frame ratio shift?
Need more seconds?
Open Gemini Omni 1.2 in Extend video mode: it adds 3–10 seconds to a clip up to 15 seconds. Describe only the next action and ask for a seamless continuation with no cut.
Video editing prompts
Keep the framing, camera movement, the woman's motion, face, hair and clothes, and the timing. Replace only the background: instead of the office, an evening waterfront with lights, as in image 1. Warm streetlight falls on her face from the left. Do not add people or text.
Keep the framing, camera movement, the man's motion, his face, hair and the timing. Replace only the jacket with the red leather jacket from image 1. Other clothes, background and light stay the same. Exactly one jacket, no new details.
Keep the composition, camera movement, people's motion and timing. Turn the whole clip into hand-drawn 2D animation: clean outlines, flat colors, soft shadows. Faces and clothes stay recognizable, the number of people and objects stays the same.
Seamless continuation with no cut. The man puts the cup on the table and turns to the window. Camera, light and framing are the same as at the end of the source video.
Price in coins
AI for video editing: what each model can do
Two models edit a finished clip. Seedance 2.5 takes a longer clip and more references; Gemini Omni is strongest at editing and can extend an uploaded clip.
| Model and mode | Source video | What to add | Result |
|---|---|---|---|
| Seedance 2.5, Video editor | One video of 4–30 s, MP4 or MOV | Up to 30 photos and up to 10 audio files | Same length and format as the source |
| Gemini Omni 1.2, From video | One video up to 15 s | Up to 10 images | Edited clip |
| Gemini Omni 1.1 and 1.0, From video | A segment up to 10 s from a clip up to 30 s | Up to 5 images | Edited segment |
| Gemini Omni 1.2, Extend video | One video up to 15 s | Description of the next action | Plus 3–10 s to the clip |
A finished Veo 3.1, Grok Imagine or Midjourney Video result is extended with the Extend button on the result when it is offered.
Replace the background in a video with AI
Describe the new place and immediately list what must not change: the person, their motion, clothes and the camera. Add a line about light — where it hits the face in the new place — or the person looks pasted in.
If the background goes wrong, prepare a photo of the new place in an image model such as Nano Banana Pro and attach it to the edit as a reference. Repeating a ready picture is easier for the model than building a place from a description.
Change clothes in a video with AI
Clothes and objects are best replaced from a photo: upload a shot of the new jacket, bag or package and name it in the prompt. State the count — exactly one bag — so the model duplicates the item less often.
An object that is barely visible in the source or keeps getting covered by a hand is replaced worst. For such shots pick a segment where the item is clearly visible.
The person stays, everything around changes
To keep the person as is and change the background, clothes and objects, list everything that belongs to them in the prompt: face, hair, motion, expressions, voice. Add do not beautify the face.
One change per pass is more reliable than three at once. Change the background first, check the result, then do the clothes in the next pass.
AI video editing for long clips
Edit a long clip in parts: pieces of 4–30 seconds for Seedance 2.5 or up to 15 seconds for Gemini Omni 1.2. Cut at shot changes, attach the same reference photos and the same change text to each piece, then join them in an edit and check the seams. The price of each part is shown before you start.
Which model to use
FAQ
Which AI can edit a finished video?
In Mixer AI it is Seedance 2.5 in Video editor mode and Gemini Omni in From video mode. Both change the background, clothes, objects and style while keeping the source motion and camera.
Can I replace a video background without changing the person?
Yes, if the prompt lists what stays: face, motion, clothes, camera and timing. A photo reference sets the new background more reliably.
How long can the video I edit be?
4 to 30 seconds in Seedance 2.5 and up to 15 seconds in Gemini Omni 1.2. Gemini Omni 1.1 and 1.0 edit a segment up to 10 seconds from a clip up to 30 seconds.
Can AI extend an uploaded video?
Yes, Extend video mode in Gemini Omni 1.2 adds 3–10 seconds to a clip up to 15 seconds. Veo 3.1, Grok Imagine and Midjourney Video results have an Extend button when it is offered.