Over the last few weeks, we’ve been building RateMyVid.ai — a platform where you upload a video, the AI watches it, reacts to it, and gives it a humour score.
The original idea started as a joke.
People already send videos to friends asking:
“Is this actually funny?”
“Would this go viral?”
“Be honest…”
So we thought:
What if AI could judge it instead?
But behind the scenes, this became something much bigger for us.
We’re currently developing OttBot Vision, our video understanding technology, and we wanted a project that could genuinely stress test it against unpredictable real-world content instead of curated demo footage.
Funny videos turned out to be the perfect challenge.
Humour is chaotic.
It depends on:
timing
speech
context
visuals
reactions
randomness
Which makes it surprisingly difficult for AI to understand consistently.
The pipeline behind RateMyVid currently:
extracts speech using AI transcription
samples unique frames from the video
uses a Vision LLM to analyse scenes and context
generates commentary and reactions
produces an overall humour score
Sometimes the AI is accurate.
Sometimes it’s brutally harsh.
Sometimes it completely misses the joke.
Honestly, that unpredictability became part of the fun.
A lot of AI products right now focus on productivity and optimisation. We wanted to build something lightweight, entertaining, and social, while still pushing the boundaries of what our video AI stack could do.
Still early.
Still experimenting.
But it’s been one of the most fun products we’ve built so far.
Mark — love the concept and the honesty about it being "still early, still experimenting." At $150/mo Stripe verified you have real people paying to know if they're funny — that's a stronger signal than most tools get. The question now is how many creators land on the page, laugh at the idea, and leave without paying. That drop is almost always one friction point — could be the pricing, the CTA, or something that breaks the "is this legit?" trust check. I do conversion audits — want me to find it?