If you can’t beat them, join them. I’ve spent the past few months diving into ComfyUI, a Python-based node system that works with multiple open-source LLMs. Right now, I can accomplish nearly everything the paid platforms offer (with the exception of Sora2), and I’m absolutely loving the possibilities this tool opens up. My current GPU is an NVIDIA GeForce RTX 4070 Ti—a decent card, but limited with only 16GB of VRAM. That means I’m restricted to generating images for now, not full video. With a 4090 and its 32GB of VRAM, I’d be pushing the limits of broadcast graphics and practically ruling the airwaves! Saving up is tricky with elementary-aged twins, but I’m working toward it. In the meantime, I’ve been exploring AI workflows that don’t just “create something from nothing.” I’ve started training LoRA models within LLMs—Qwen has impressed me for text, while Flux shines at generating people. My larger goal is to train AI on my station’s logo animations and Storm Tracker system, so eventually I can prompt it to generate any type of commercial scene I need—with the AI already knowing what to do.