The "15 seconds is real inventory" framing is right. That length finally matches what ad and social buyers actually buy.

I would only add: do not put every shot on Omni just because the rate dropped. I still route Veo for short photoreal dialogue, Kling for motion-heavy people shots, Seedance 2.5 when a reference pack drives the take, and Omni 1.2 Flash when I expect multi-turn edits. Treat sticker $/s as one input. Cost per usable clip after rerolls is the one that hits the budget.

Disclosure: I run PixelDojo and shipped Omni 1.2 Flash there. Bias noted. Your startup playbook still maps cleanly onto how we route shots.