AI Conference Summarizer: Never Miss a Talk You Did Not Attend

An AI conference summarizer digests hours of talks into key points. WhisperX + LLM pipeline for transcripts with speaker labels.

💡 What You Will Learn

An AI conference summarizer digests hours of talks into key points. WhisperX + LLM pipeline for transcripts with speaker labels.

Conference season means dozens of talks and zero time. An AI conference summarizer turns hours of recordings into a readable digest - and the open-source route costs nothing.

The Recipe

whisperX (23,466 stars) transcribes with word-level timestamps and speaker diarization, so you get who-said-what. Feed the transcript to an LLM via Ollama (177,874 stars) with a per-talk prompt: core claim, evidence, tools mentioned, one-sentence takeaway. yt-dlp (182,988 stars) pulls the talks from YouTube automatically - the whole pipeline runs on a cron job.

What You Get

A digest per talk in under 5 minutes of processing: 3-5 key points, quotes with timestamps, and a link back to the exact moment. That timestamp link is the killer feature - it makes the AI summary a navigation tool, not a replacement for watching.

FAQ

Q: Does it work with slides?
A: The audio pipeline ignores slides; pair it with MinerU (77,061 stars) if you have the deck PDFs for slide-level extraction.

Q: How long to summarize a 40-min talk?
A: Roughly 3-5 minutes of compute on CPU; under a minute with a GPU.