Just shipped · 2026-05-24

Google Omni is now live for everyone.

Gemini Omni rolled out today to all Google Flow users. Natural-language video editing, two free generations a day, no waitlist.

"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:

Free tier · Now live
2 generations / day
Non-subscribers, no card, no waitlist. Subscribers keep their existing quota.

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:

// excitement
"This unlocks ad iteration speed I didn't think we'd see until 2027."
// caveat
"2 gens/day is a tease — most ideas don't survive first contact with one render."
// friction
"Policy filter is heavy. Half my prompts came back rejected on first try."
// comparison
"Seedance still wins on cinematic camera moves. Omni wins on text-in-scene."

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.