Members-Only
Recent Talks & Demos are for members only
You must be an AI Tinkerers active member to view these talks and demos.
January 30, 2025
·
San Francisco
Instantcasts: Fast Whisper Transcripts
Learn how to process an hour‑long podcast in under ten seconds using optimized Whisper inference, covering model tweaks, audio storage, chunking, prompting, and diarization.
Overview
I’ve been working on a fun side project around fast Whisper inference that takes the URL to a podcast and, in <10 seconds for an hour-long show, generates a transcript and summary. The actual application is super basic, but it showcases some advanced stuff around optimizing Whisper inference at both a model level and an infra level (e.g. where does the podcast audio file live? it matters!)
Tech stack