The news | Longer clips arrive alongside continued editing

On November 5, 2026, Google released Gemini Omni 1.2 Flash, generally available through the Gemini API and Google Cloud. Omni is an AI model for generating and editing video. The new version raises the native clip limit from 10 to 15 seconds and supports follow-up conversational edits to existing video. Google also describes generating a lower-resolution draft before upscaling a chosen result. Official announcement

Why it helps | Ask to preserve the subject while changing a detail

Suppose an existing clip introduces a handmade cup. The subject and camera movement are right, but you want a lighter background. Explicitly ask to preserve the cup’s appearance and motion while changing only the background, then inspect the rim, pattern, and continuity. Multi-turn editing offers a way to request this, but each result still needs review; consistency in a vendor demonstration is not a guaranteed outcome.

Model inputs are not necessarily available through every entry point for every account. For example, the Google Cloud announcement still requires allowlist access for audio references. This background edit does not need an audio reference. Cloud availability

How to use it | Edit an existing clip in Flow

The following uses Google Flow. Video editing has country or region restrictions. Editing a 15-second clip requires Google AI Plus, Pro, or Ultra; other eligible accounts are limited to 10 seconds. Editing consumes the same credits per second as generation.

  1. Choose the model. Open an existing Flow project, open generation settings beneath the input box, choose Video, then Omni Flash 1.2.
  2. Select a clip. Open the video and select a segment within the allowed duration in the trim window. Start with a short clip already in the project.
  3. Describe one change. Enter “Change the background to pale beige; preserve the cup pattern and camera movement,” click Generate, and compare the subject and background with the original.
  4. Continue or return to an earlier version. Add a follow-up request if needed. The help page describes retaining editing context for up to three conversational turns. The original remains available; History shows earlier versions and their prompts. Flow instructions