Your AI demo looks impressive on first glance, but that's kind of the problem: it feels like a demo, not a meltdown.
The uncanny gap isn't in the tears, it's in everything that usually surrounds them – messy breathing, broken pacing, the way people avoid eye contact with the lens when they're actually not okay.
When AI can generate those invisible "I shouldn't be posting this" signals on demand, then the VV curve will really change.
Curious: have you A/B tested your AI crying clips against real emotional ones on cold audiences yet?
This thread perfectly explains why most brands are stuck at 20–30% VV and "UGC fatigue":
they're optimizing scripts, not states.
A crying reaction isn't a "format", it's a nervous system in fight-or-flight on camera.
You can copy the hook, the lighting, even the tears with AI – but if you don't copy the stakes, viewers subconsciously clock it as safe and scroll.
Creators who understand this will eat everyone trying to out-edit their way to retention.