Michael Rea
Michael Rea10mo ago

XO Brain Ai

https://xobrain.ai/ https://devpost.com/software/xo-mf458w My hackathon entry building an External Brain Ai. I plan to push forward beyond the hackathon and keep working on this.
Devpost
XO Brain Ai
XO Brain is your AI external brain. Effortlessly capturing and organizing your thoughts, notes, and tasks in real-time. XO turns your chaotic stream of consciousness into structured notes.
No description
No description
No description
No description
5 Replies
KinKon
KinKon10mo ago
Cool project
Hardik Verma
Hardik Verma10mo ago
@Michael Rea Nice project, are you splitting the chunks while recording and sending it to something like whisper?
Michael Rea
Michael ReaOP10mo ago
I tried to use whisper at first but knowing when to split the audio was tricky and latency was too high. I might still use whisper for a final Passover at the end to correct potential transcription errors. I can then use the word level timestamps to have a playable audio snippet per chunk that's in the note. I'm using deepgram to do the splitting while recording, it easier waits for a small gap in your speech to then send to my API for processing. The ms gap in speech still needs fine tuning but I've leaned towards higher so you get bigger chunks sent through.
Michael Rea
Michael ReaOP10mo ago
Deepgram Docs
Transcription features - Endpointing - Deepgram Docs
Endpointing returns transcripts when pauses in speech are detected.
Hardik Verma
Hardik Verma10mo ago
Great thanks, i am also building something similar for learning purposes only where i am sending the audio directly to whisper to generate transcription and then using mistral on together ai to generate data out of transcript but the turn around time is huge as whisper is too slow.

Did you find this page helpful?