Models · 14 June 2026 · 2 min read
Gemini Omni Flash: Google’s video model from I/O
Announced at Google I/O on 20 May 2026: video made and edited from mixed inputs, conversational edits and a watermark on everything.
By The Crewnema team
Gemini Omni is a family of models from Google DeepMind that makes and edits video from any mix of images, audio, video and text. The first one, Gemini Omni Flash, was announced at Google I/O on 20 May 2026. This note sticks to what Google itself said.
What Google announced
- Video made and edited from a mix of images, audio, video and text in one request.
- Conversational editing: each instruction builds on the last, and Google says characters and scene context stay consistent across the edits.
- Better handling of physics such as gravity and fluids, helped by the wider Gemini model’s knowledge of the world.
- A SynthID watermark on every generated video by default, so it can be checked later.
- Rollout to the Gemini app and Flow for paid subscribers, and to YouTube Shorts and the YouTube Create app at no cost. Developer access was to follow in the coming weeks.
Google also said what it was holding back. General audio and speech editing was kept back for now while it worked out how to offer it responsibly, because changing someone’s voice without their consent is an obvious risk. Initial clips were capped at 10 seconds, which Google described as a deployment choice and not a technical limit.
What it means for an advert
Conversational editing matches the most common request after watching a draft: keep all of that and change one thing. A watermark on everything is worth knowing about if you publish AI video, because viewers and platforms can check it. It is also a reminder to say plainly when a video is AI-made where a platform asks you to.
A later update added extension, first and last frames and draft resolutions: see Gemini Omni 1.1 Flash. Prompts for any of these models work best when they describe one action in one place; see seven prompting techniques.
Sources
Make your first video
Crewnema is in private beta and lets businesses in a few at a time. Request access and we will write when yours is open.