adds ~600 videos to pull from YouTube and test the latest Whisper Turbo model on them #djangotv

⬇️ download another 100+ conference videos and transcribe them via mlx_whisper #djangotv

Trying insanely-fast-whisper ⚡️ for fun #chores

🤖 got sweep running on my Mac Studio and spent $0.73 letting it write five pull requests to try it out. Not bad.

Play around with Whisper

🤖 use the python whisper api to translate some podcasts to text files. Mostly just worked. 🤷

🚜 wrote an app to pull the latest podcast audio files so they may be processed with Whisper (I prefer MacWhisper) to build out transcripts of my favorite episodes #mylife

#thecompaniesapi work on performances PR ; watch the GPU struggle with lots of queries

figure out whether Vercel is OK for longer running video editing background tasks #ship30

Over the past three days and a half, I've dedicated all my time to developing a new product that leverages locally running Llama 3.1 for real-time AI responses. It's now available on Macs with M series chips – completely free, local, and incredibly fast. Get it: Snapbox.app #snapbox

🤖 got llama-cpp running locally 🐍

research cpu gpu performance bottlenecks on xen passthru #optimizing

Read the up on MLX for Apple Silicon.

Get out of sync audio after rendering overnight from graphics-heavy part onwards, so attempt rendering again or reduce gfx💀 #vids

Lunch & watch WWDC vids 🧀👨‍💻

do thread how to play video games on MacBook M1/M2 #life twitter.com/levelsio/status/1…

get 108k+ runs on my public speech to text pipeline replicate.com/thomasmol/whisp…

test fine-tuned whisper #spectropic

📼 finally figured out how to make VHS not blow up on macOS. I can finally script and record terminal demos.

🖥 ordered a 32GB Mac Mini Pro so I can do more AI work and have a dedicated machine at home again #mylife