"It's like Nano Banana, but for video." That's how the Google Flow team framed Gemini Omni when they teased it earlier this week. Today the gate dropped: everyone gets in.
The pitch is simple. Type what you want to change in a clip. Omni does it. No keyframes, no masks, no second pass. The demo posted by @FlowbyGoogle shows the model embedding the words "GEMINI OMNI" into beach scenes naturally — woven into towels, printed on umbrellas, baked into picnic blankets. One prompt. Ten seconds. Done.
01 / The releaseWhat changed today
Until this morning, Omni was gated behind a waitlist and a Google AI Pro / Ultra plan. The announcement post flipped two things at once:
02 / Why it mattersThe "Nano Banana" moment for video
Nano Banana was the inflection point for image editing — natural-language edits that actually preserved subject identity and physics. Omni is Google's bet that the same trick scales to motion. Early outputs suggest the temporal consistency is real: text stays attached to surfaces frame to frame, lighting holds, no flicker on the embedded elements.
"Two free gens a day means every creator on the planet can try this before lunch tomorrow. That's not a feature flag flip. That's a market move." — Reaction thread, 17h ago
03 / The room's readWhat people are saying
Reception across the post is split three ways:
04 / What I'm doing with itThe video I'm cutting next
Making a teardown. Same prompt, three models — Omni, Seedance, Runway Gen-4. Same 10-second target. Cost, quality, latency, prompt adherence. Drops on the channel this week. /subscribe-if-you-haven't.
If you have a prompt you want me to throw at it, ping me on X. Best three get tested live.