Cinemas spend millions on marketing. But your face doesn’t lie. I built a $10 hardware system to prove it.
okay so this started as a stupid weekend project and now i cant stop thinking about where it can go
here’s what i built
my laptop runs face_recognition, pulls face mesh landmarks, figures out if you’re happy or not, and sends that over HTTP to my ESP8266 bot. the bot gets the emotion, shows a reaction on the OLED, and logs it. that’s it. simple.
but the architecture is the interesting part.
ESP8266 does nothing heavy. it just receives, reacts, logs. cheap as chips. the laptop (or any server) does all the thinking — detection, mesh, classification. thin client model basically. which means you could deploy 50 of these nodes for almost nothing, all talking to one central machine.
so then i thought
what if you put these in a cinema?
nodes on every row. audience faces tracked during the film. emotion data collected per scene, timestamped. at the end you don’t ask people what they thought. you already know. you have the data from their actual face.
honest review generated by facial reactions. not by critics, not by surveys, not by paid promotions.
this is very early stage. the bot works, the loop works, but scaling this is a different problem.
need your brains on this
• how would you handle the compute side at scale — local server or cloud?
• better real-time emotion classifiers?
• privacy — how do you ethically collect face data in a public space?
• anyone done something similar?
drop ideas below 👇