Real-world
Media
Concrete workflows, including the tradeoffs and limitations reported by their authors.
this collection
Describing security camera events on an 8GB card
A vision model on a modest 3070 turns raw detection events from a home-built surveillance system into plain-language descriptions.
A dedicated local editing model for novels, running on two 5060 Ti cards
A writer runs Qwen 3.8 27B locally on two RTX 5060 Ti cards with a model card fine-tuned only to edit books and detect plot holes, not to write.
Iterating on generated video locally instead of burning generation credits
A user generates video locally with a model like Minimax H3, trading longer render times for unlimited retries and no per-generation costs.
Local models as checker and tracker for a text adventure engine
A game developer on an RTX 5080 uses Gemma 4 26B for background passes in a text adventure engine, while larger models handle knowledge-heavy generation.
Generating product videos with ComfyUI workflows on an M1 Max
Thirty to forty custom ComfyUI workflows generate product videos overnight on an M1 Max, using a photographed person as a virtual product model.
Local Telegram-to-comic pipeline on an NVIDIA Jetson Thor with Qwen 3.8 27B, Qwen-Image 2.1 and ComfyUI
A 128GB NVIDIA Jetson Thor T5000 runs Qwen 3.8 27B, Qwen-Image 2.1, ComfyUI and the Hermes agent to turn one Telegram message into a comic book and deploy it to a website, entirely locally.
AI tagging for a media library on an 8GB card
Proving out AI tagging and logging for a media archive on one of the smallest cards in the thread, then building toward a self-hosted digital asset manager.