Hey All - Wanted to share an idea that's been in my brain for a while now but finally had a chance to get setup an MVP.
Everyone's chatting with one AI, and then when voice got integrated, everyone lost their minds. OK, cool. But what really makes things come alive is when you have more than one AI.
What if you could have a group of AI friends, and you could just chat with them?
In this version, I’ve got two personalities—Jake and Kathie—who can have a conversation with me. Here’s the rundown:
- Speech Transcription: I’m using browser-native functionality to transcribe my speech (been having issues with using third party services and web sockets)
- LLM Processing: The transcription is sent to an LLM via Groq, where it generates responses and assigns the speaker based on Jake and Kathie’s personalities.
- Text-to-Speech: The responses are then turned into speech using Deepgram’s TTS engine.
It’s a basic setup for now, but it’s already showing a lot of promise. There are a few bugs, like some voice leakage between the AI personalities and wanting them to interrupt each other naturally, but I’m really excited about where this is headed.
Looking for feedback and ideas on how to take this further! Here’s a video demo (turn up your volume a bit—my Macbook speakers were a little far away when I recorded it).
[link] [comments]