AI video editing workflow: from reference to finished clip

An AI video editing workflow in Flow AI Video starts with a clear shot brief and ends with reviewing a generated clip in Create. For an existing video, use a reference-based generation to explore a visual change. This is different from cutting clips on a frame-accurate timeline.

How do I choose the right input?

Start with text for a new scene, an image for a visual starting point, or a reference video for a motion example. Use the AI Video Editor when your goal is to regenerate a scene or style from footage you already have.

Keep the reference focused on the subject you want to change. A cluttered clip makes it harder to explain which action matters. Write what should stay recognizable as well as the change you want.

How do I prepare and submit the shot?

Select a model first, then review its available aspect ratios, duration, and resolution. For image-to-video, use the Start and End slots deliberately; an end image belongs in End even if it is the only image you upload. A video reference belongs in the video attachment area.

Try: “Keep the person and walking direction recognizable. Change the setting to a quiet street at dusk, with warm window light and a steady side-tracking camera.” The prompt expresses a direction rather than guaranteeing that every detail will be preserved.

Check the estimated credits and click Generate. If your balance is insufficient, use the dialog to compare plans or keep editing. A submitted request opens in Create with its input media and prompt.

What should I review before downloading?

Watch the entire clip. Compare the subject with the reference, check the beginning and ending, and inspect text, packaging, hands, and unexpected movements. For another take, change one instruction so you can tell what affected the result.

Download a usable clip and finish captions, precise cuts, or final audio in the editing workflow that suits your project. This guide does not imply those controls exist inside the reference-generation composer.

Prepare a video edit or compare image-to-video inputs.