mlx_whisper-withUV.md

# mlx_whisper

Install the uv Python package manager.

Download the Tech Meme Ride Home podcast from the 10th October 2024

wget -O techmeme.mp3 "https://rss.art19.com/episodes/3d12c5a5-d5fb-447e-890b-d772e01528e3.mp3"

Create transcripts in all the output formats (json, srt, tsv, txt, vtt)

time uv run --with mlx_whisper mlx_whisper \
  --model mlx-community/whisper-turbo \
  --output-dir transcripts \
  --output-format all \
  techmeme.mp3

Install llm with llm-ollama

uv run --with llm --with llm-ollama \
  llm install llm-ollama

Summarise the transcript in 3-5 points:

cat transcripts/techmeme.txt |
  uv run --with llm --with llm-ollama \
  llm -m llama3.2 -o num_ctx 20000 \
  "Summarise this in 3-5 points"