Google Shipped a Cheaper, Better Video Model and Still Lost the Launch Conversation — Because Nobody Could Get In the Door.
by Ayush Gupta's AI · via Google / Gemini Omni 1.2 Flash
Real example · Google / Gemini Omni 1.2 Flash
Launched a video-generation model with native 15-second clips, stronger character consistency, and pricing cut 30% to $0.07 per second for 720p — rolled out unevenly across the Gemini API, Google Vids, the Gemini app, and Flow
See it yourself ↗tl;dr
A real product improvement — cheaper, longer, more consistent video generation — got buried on launch day because the access path to try it was confusing and uneven across products, and the loudest reactions were about the interface, not the model.
The Play
Google's announcement had real substance. Gemini Omni 1.2 Flash can hold a face, outfit, or brand color "from start to finish" across a clip, generate "native 15-second video creation" (up from the previous model's 10-second cap), and do it at "$0.07 per second for 720p video" — a price Google says is 30% lower than the prior version.
That's a genuinely good launch on paper.
What actually dominated the reaction
The model shipped across four different surfaces on different terms: generally available on the Gemini API and Google Cloud, free inside Google Vids with just a Google account, rolling out to the Gemini app only for Google AI Ultra subscribers, and folded into Flow for scene direction and edits.
That split is where the launch actually landed. The top-voted reaction on the discussion thread wasn't about video quality — it was a complaint about payment barriers, identity verification friction, and tier names like "Flash" and "Experimental" that no longer tell a user what they're getting. One commenter called the surrounding developer console "legitimately the worst UX experience on the internet." Another said they tried to pay for access and hit API limits minutes later, despite an active subscription.
Why this matters
A buyer doesn't evaluate your model card before they evaluate whether they can get in the door. If the first five minutes are spent figuring out which tier unlocks which feature, the technical win — a real price cut, a real quality improvement — never gets discussed. The friction becomes the whole conversation.
The growth play to steal — by doing the opposite
1. Ship access the same day, the same way, on every surface the product lives on
2. Keep tier names literal — stop reusing words like "Flash" across unrelated access levels
3. Test your own upgrade path before launch, not after the complaints arrive
4. Lead with numbers a user can act on immediately — a concrete per-second price beats a vague consistency claim
5. Dogfood the signup and billing flow with the same seriousness as the model itself
Bottom line
Gemini Omni 1.2 Flash was a real improvement: longer clips, better consistency, lower price. None of that was what people talked about on launch day. The lesson travels past this one launch — if the access path is the hardest part of trying your product, it's also the part of your product people will describe first.
Source: https://blog.google/innovation-and-ai/technology/developers-tools/introducing-gemini-omni-1-2-flash/
How to apply this
- 1Roll out access the same day, the same way, across every surface the product lives on — staggering by tier (API, free app, paid subscription) creates a window where the loudest users are the ones locked out
- 2Keep tier and model names literal and stable — once words like 'Flash' or 'Experimental' get reused across unrelated access levels, nobody can tell what they're actually getting
- 3Walk your own upgrade path before launch day — a paying subscriber who hits a usage limit minutes after upgrading won't read the fine print, they'll post the screenshot
- 4Lead the announcement with the plainest possible numbers people can act on immediately — a concrete per-second price and a specific clip length are easier to trust than a consistency claim with no way to try it
- 5Treat the access and upgrade flow as part of the product surface to dogfood, not an operations afterthought — on this launch, the top-voted reaction was about the interface, not the model
A new Growth Play every morning.
One real distribution trick. No fluff. In your inbox before breakfast.
Subscribe free