Skip to content
AI Primer
workflow

Gemini Omni Flash supports targeted frog-to-kitten video edits in creator demos

Two creator posts show Gemini Omni Flash altering a specific subject inside a clip while leaving most of the surrounding motion intact. That matters because object-level video edits appear usable for targeted fixes instead of full rerenders.

3 min read
Gemini Omni Flash supports targeted frog-to-kitten video edits in creator demos
Gemini Omni Flash supports targeted frog-to-kitten video edits in creator demos

TL;DR

Google's official announcement promises conversational edits that preserve continuity, and the creator demos here get unusually close to that sales pitch. You can also watch AmirMushich's Live mode example, skim Google's Flow rollout notes, and jump to ozansihay's longer tutorial for the rough edges, especially around audio and end-to-end workflow.

Frog-to-kitten swaps

The cleanest reveal in this batch is subject isolation. In venturetwins' clip, a frog becomes a tiny kitten while the framing and the rest of the shot barely move.

ozansihay's first-pass edit lands a similar result on a different source clip, which matters more than a single cherry-picked sample. Google's Gemini Omni post describes this as conversational editing that keeps character, physics, and scene coherence across revisions.

Live camera prompts

The app surface is as interesting as the model behavior. AmirMushich's Live mode clip shows Gemini taking a live camera view, then applying spoken edit requests inside the same session.

That matches Google's Gemini app update, which folds Gemini Live deeper into the app, and Google's Omni launch post says Omni Flash is rolling out to the Gemini app, Google Flow, and YouTube Shorts.

Access and costs

The official rollout is split by surface. Google's Gemini Omni announcement says Omni Flash is available to Google AI Plus, Pro, and Ultra subscribers in the Gemini app and Flow, while YouTube Shorts and YouTube Create get it at no cost.

YutoIT's breakdown adds the practical numbers missing from the launch post:

  • Text-to-video runs about 15 credits for 10 seconds.
  • Video-to-video editing deducts credits based on the uploaded clip.
  • 4K export costs 50 extra credits per download.
  • Default output is 1080p.

ozansihay's longer tutorial also calls out a limit that the short demos hide: audio continuity still needs cleanup, even when the visual edit lands cleanly.

Share on X