Inspiration

In 2022, my (phuthai) mother passed away as I was graduating RPI (now doing masters). Kelly and I were inspired to use AI to help re-live interactions with past (or current) loved ones that are no longer near. Keeping with "Retro v Modern", my mother and many people in our age group have parents who were living it up "retro" 80s. We put a happy spin and and theme on this interaction, and hopefully bring joy/entertainment to anyone who wants a retro version/conversation with a person.
NOTE: IF YOU VISIT THE URL, YOU MUST BE ON RPI WIFI. I BLOCKED SUBNETS 4 SECURITY

What it does

Upload an image of anyone, and watch them go 80s!
Pre-Prompt the interaction, and carry on unforgettable conversations in a variety of voices!
Render an interaction, and watch as the image comes to life!

Because we ran our own local machine learning model to create lip-syncing images, the Python required heavy CUDA/CPU usage. We had to use the more expensive VULTR VMs to continue running remotely
Because of computational requirements, I neglected to perfect the Kubernetes deployment. The Cluster is up, but the app is not functional on Kubernetes
Web Cam doesn't work when run on the server, only locally
There are 2 versions of Google Gemini's APIs, that aren't compatible. Using AI to learn this API causes issues since even Gemini mixes up the APIs. (google-genai vs google-generativeai)
Repo is private, and we'll offer on request or make it Public after the Hackathon ends. We did not do a good job cleaning up the API keys from the repo + repo history
We both hit CoPilot prompting limits