I built a TikTok machine before I ever opened TikTok
Confession: I've never posted a TikTok, and I've barely watched one (I'm a Youtube Shorts addict though 😂).
So naturally, I built a machine that makes them.
It's for thegoatdebate.io, my Messi vs Ronaldo GOAT debate site. The star is Papa Goat, a grumpy, deeply unimpressed goat who has heard your GOAT take before and didn't like it then either. 🐐
How it works: you give it some football footage. It picks a take from a bank of about 150 sourced facts, and a local AI model writes Papa Goat's script. ElevenLabs gives him a voice, and a talking goat lands on top of the video. Out comes a vertical video that a human (me) watches before anything gets posted.
Polar is a local, privacy-focused AI desktop assistant designed around a futuristic HUD interface and system-level interaction.
The project explored how a desktop AI could understand the user's environment, process visual and voice input, and respond or perform actions without relying entirely on cloud services.
Key features:
Local AI assistant architecture
Futuristic Tauri-based desktop HUD
Screen and contextual awareness
OCR-based extraction of text from the screen
Voice interaction pipeline
AI-powered context processing and responses
System-level desktop interaction and automation
Local/offline model execution
Real-time assistant-style interface
OCR pipeline:
Screen capture → OCR → Context extraction → Local AI → Response/Action → HUD
The project combined AI, computer vision, OCR, voice interaction, desktop application development, and modern UI engineering into a single experimental personal-assistant platform.
My contribution: Architecture, application development, AI integration, OCR functionality, UI/HUD development, and system interaction.