@Prajwal – Hack Club Scrapbook

Second demo video of my JARVIS project, with sound 👇 On May 30th, I had the opportunity of demoing JARVIS at the Vancouver AI Community Meetup! In this video, I showcase JARVIS, an AI voice assistant on a hat! • It can describe what it sees in front of you • It can search the internet and answer your questions • It can converse with you with low latency thanks to Groq's blazing fast speeds An application of this technology is for visually impaired people to know more about what they are looking at or facing, know their surroundings, and have someone to talk with and get help from! There are a lot of companies (such as Seleste) who are trying to achieve the same goal, but what I think is unique about my project is that everything I used was either already on me (Raspberry Pi) or was very cheap to buy (mic + speaker), and the software I used is completely free, making this proof of concept very economically viable and useful. Updates on JARVIS: • Added a speaker so others can hear it • Used a smaller microphone to reduce weight • Used Gemini v1.5 Flash for vision-based prompts • Used Groq's Llama 3 70B for text-based prompts and for conversation mode • Implemented conversation mode so the user can talk to JARVIS as if it were their friend • Implemented short-term memory for conversations • Added faster voice streaming • Added ability to stop speech from JARVIS by clicking the button This was my third attempt at the event to try to get it working. I'm glad I fixed the issues I encountered and that everyone loved the demo! The video was recorded from my friend's phone, so it's not of the highest quality, and a bit of the last part of the demo got cut off, but I'm happy with how it turned out! LinkedIn post | GitHub code