Been casually working on some AI projects lately. Open-sourced both: LLM Archive Downloader. Got paranoid one night about open-source models potentially disappearing (licensing changes, restrictions, who knows). So I built a script to download and archive them locally. One command, pick what you want — Llama, Mistral, DeepSeek, Whisper, the works. ~200GB total or grab the essentials (~40GB). Now they're sitting on my hard drive. Offline forever.

→ github.com/Ashwinsuriya/llm-archive-downloader

YouTube Shorts Generator Converts long YouTube videos into Shorts automatically. Whisper handles transcription, LLaVA analyzes frames to find interesting moments, Mistral picks clips and writes captions. Everything runs locally in a parallel pipeline. No APIs, no subscriptions. → https://github.com/Ashwinsuriya/yt-shorts-generator Nothing fancy.

Just scratching my own itch and sharing in case anyone finds them useful.