Skip to main content

AI Video Editor: Change an Existing Clip

Start with a clip and describe the change.

Video
Video
A warmer treatment of the park scene
A warmer treatment of the park scene

Edit @video_1: make the park lighting warmer and slightly golden, like soft warm sunlight, with gently warmer highlights and a subtle golden tint across the scene. Keep everything else exactly as in the source: the same real woman with the same face, hair and clothes, the same walking motion and timing, the same camera framing and movement, the same park layout, and the same duration. Generate quiet, natural park ambience without speech or music. No text, subtitles, captions, watermarks or logos in the frame.

3 s

How it works

Try it now
  1. 01

    Add your video

    Drop it in, or pick one of yours.

  2. 02

    Say what to change

    A sentence or two is enough.

  3. A warmer treatment of the park scene
    03

    Make it

    Start creating, then keep working with HowWorks.

Explore the possibilities

A warmer treatment of the park scene

About

Upload a source video and explain what should change while keeping the rest of the shot coherent. Generative video editing is useful for trying another appearance or atmosphere without starting from a text-only scene. Describe the intended result and the important elements to preserve, such as subject, action or camera movement. Current model controls determine the accepted references and sound options. Review the full result for consistency and changes you did not request. This page is a generative clip-editing task, not a timeline editor.

Questions

How is this different from text-to-video?

Video editing starts with an existing clip. Text-to-video creates a new shot from a written description.

Can I preserve the original sound?

Choose a compatible model and its original-audio option when it is available. Check the result rather than assuming every model preserves sound.

Will only the requested detail change?

Generative editing can alter other details. Describe what matters and compare the full output with the source.

From the blog