Fork of matthiasn/talk-transcripts. Thanks to Matthias Nehlsen for the original collection. This fork adds automated transcription tooling using Whisper and ffmpeg.
Transcripts of technical talks, organized by speaker. Designed for reading — because some of us learn better from text than from video.
| Directory | Talks |
|---|---|
| Hickey_Rich | 43 — Simple Made Easy, Are We There Yet, Hammock Driven Development, ... |
| Szulc_Pawel | 8 — Free Monads, Tagless Final, Curry-Howard, Lenses, ... |
| Halloway_Stuart | 7 |
| Nolen_David | 3 |
| Kay_Alan | 2 |
| Miller_Alex | 2 |
| Sierra_Stuart | 2 |
| Steele_Guy | 2 |
| Bjarnason_Runar | 1 — Constraints Liberate, Liberties Constrain |
| Czaplicki_Evan | 1 |
| DeGoes_John | 1 — One Monad to Rule Them All |
| Jacobowitz_Luka | 1 — Monoids, Monoids, Monoids |
| Labella_Fabio | 1 — How do Fibers Work |
| Norris_Rob | 1 — Functional Programming with Effects |
| Trunk_James | 1 — Clojure in a Nutshell |
The tools/ directory has a complete pipeline for adding new transcripts from YouTube:
- Download & transcribe audio with faster-whisper (OpenAI Whisper)
- Extract slides via ffmpeg scene detection
- Post-process — fix technical terms Whisper gets wrong (Dijkstra, Gödel, Curry-Howard, ...), merge segments into paragraphs, align with slides
cd tools
uv venv --python 3.13 .venv && uv pip install -e .
.venv/bin/python batch_transcribe.py sources.txt -o ../Hickey_Rich/ --speaker "Rich Hickey" --device cpu --compute-type float32
.venv/bin/python extract_slides.py sources.txt -o ../Hickey_Rich/
.venv/bin/python postprocess_transcript.py ../Hickey_Rich/ --in-placeSee tools/README.md for full documentation.
Please contribute transcripts and fixes to the upstream repo: matthiasn/talk-transcripts. If you read a transcript and find a URL mentioned in the talk, submit a PR turning it into a markdown link. If you transcribe a new talk, send it there too.
For more on the original project: blog post.