Writing
Essays, experiments, and notes on software, AI, and building things.
A walkthrough of how we rebuilt a live lecture transcription system — same goal of capturing only the professor's voice, but with a smarter approach: transcribe everything first, filter by timestamps after.
A walkthrough of building a real-time audio transcription system that captures only one target speaker from a noisy room — and ignores everyone else.
Learn how to stream live audio from the browser, filter silence with VAD, and transcribe speech in real-time using FastAPI, WebSockets, and OpenAI Whisper.
What actually is LSTM and how it differs from RNN network and how it can be used to identify human speech out of the background noise