AI Image Generation Platform
Stable Diffusion · Fooocus · Model Library
Tech Stack
Introduction
Welcome to the ultimate playground of creativity! This is the section where imagination bridges the gap directly into computation. I extensively push the boundaries of Stable Diffusion and Fooocus to create vibrant, intelligent art and experiment with generative workflows.
It features hands-on tooling integrations, a custom macOS-optimized backend, and centralized model management. Whether running completely natively on Apple Silicon or dockerized, it serves as my complete, high-performance toolkit for AI synthesis.
Quick Start & Architecture
The platform supports both completely containerized execution via Docker and bare-metal optimized execution on Apple Silicon.
Model Management
To avoid duplication, the installation uses a centralized storage approach with symbolic links. Checkpoints and LoRAs are kept in dedicated libraries and mapped into the application at runtime.
Fooocus/models/checkpoints → ../../models Fooocus/models/loras → ../../LoRAs
macOS & MPS Acceleration
Automatically detects macOS environments and configures PyTorch with Metal Performance Shaders (MPS) for GPU acceleration on M-series chips (Apple Silicon M1/M2/M3/M4). It enables aggressive VRAM optimizations like --all-in-fp16 and --always-high-vram for peak token-generation speed.
Curated Model Library
The platform comes heavily armed with a massive 46GB library featuring 7 SDXL Checkpoints and over 40 LoRAs downloaded from HuggingFace and CivitAI.
Featured Checkpoints
Juggernaut XL v8
SDXL 1.0
General purpose, high quality realistic model
Anima Pencil XL v5
SDXL 1.0
Anime and detailed illustration style
Realistic Stock Photo
SDXL 1.0
Stock photography realism aesthetic
SDXL Ronghua v4.5
SDXL 1.0
Chinese aesthetic and style (国风)
Selected LoRAs
Expert Tips & Advanced Workflows
Beyond standard generation, I also document advanced techniques and highly specialized workflows. Read through these guides to understand the finer details of AI creation:
Flux Prompt Crafting
Learn to systematically construct highly detailed, natural language prompts optimized specifically for the unparalleled coherency of Flux models.
Kling AI Video Directing
Step-by-step techniques for translating static ideas or image generations into dynamic, cinematic video sequences using Kling AI.
CivitAI LoRA Training
A comprehensive teardown on how to gather datasets, configure epochs, and successfully train custom Character or Stylistic LoRAs on CivitAI.