🎨 Image Generation

What it is

A local image generation pipeline built on ComfyUI, running GGUF-quantized diffusion models from City96. Designed to fit on a single consumer GPU (GTX 1080 Ti 11GB VRAM) — no cloud APIs, no data leaving the machine.

Stack

• ComfyUI — node-based workflow engine for diffusion models
• FLUX.1-schnell — primary model, GGUF quantized for VRAM constraints
• City96 quantizations — GGUF variants of mainstream models that fit consumer GPUs
• Offline-first — models downloaded once, all generation happens locally

Hardware

Why it matters

Most AI image gen assumes you have an A100 or will pay for API calls. This setup proves quality generation is possible on aging consumer hardware — just requires smart quantization and careful memory management. The models are gated on HuggingFace, so you need a token, but once downloaded, everything runs offline.