The "Big One" v1: Modular Illustrious Generation & Refinement

This is a complete ground-up rebuild of my daily driver. While it’s not the most "massive" workflow on the block, it’s built for heavy-duty precision and high-quality output without the clutter.

Key Features:

Modular Prompting Architecture: Uses a triple-layered positive prompt system (Subject, Pose, and Scene). These are automatically concatenated, allowing you to swap out environments or poses instantly without digging through a single wall of text.

Dual-Stage Generation & Upscaling: Built with a dedicated second checkpoint loader specifically for the upscaler. This allows you to generate with one model and use a different, more detailed model for the PixelKSampleUpscaler pass.

Smart LoRa Management: Features independent LoRa stacks for both the base generation and the upscale stage. The workflow is designed to automatically extract and inject LoRa trigger words directly into your prompt, ensuring the model actually "sees" what the LoRa intends to add.

Integrated Face Detailer: Includes a custom FaceDetailer subgraph that utilizes UltralyticsDetector and SAMLoader to automatically detect and polish facial features in the final pass.

Advanced Sampling Logic: Leverages DynamicThresholding and the Detail Daemon to maintain high contrast and intricate details, even at higher CFG scales.

Note: Because it pulls trigger words automatically, just keep an eye on potential prompt conflicts if you’re stacking multiple heavy LoRas!

I use a custom built Gemini gem to create my prompts. This gem lets me input a description, or a picture, and get a prompt.