Watch the full Video Tutorial here: https://youtu.be/he2lkOzZ2Io

This workflow has been designed to combine maximum artistic control with a clean, high-performance interface. It serves as a stable foundation for SDXL generations and follows the same logic as our upcoming Flux and Z-Image modules.

Core Components & Architecture

• Central Subgraph System: All complex calculations are offloaded into subgraphs (groups) to avoid "spaghetti node" layouts. The main control is handled via a clear dashboard.

• Prompt Management:

  • Portrait Master v2: Precise control over pose, facial features, and lighting.

  • Foocus Styler: Integration of over 100 style presets for instant artistic looks.

  • Prompt Mixer & Concat: A logic unit to combine standard prompts and styler outputs, including weight control (order determines influence).

• Sampling & Detail Control:

  • Lying Sigma Sampler: Specialized fine-tuning of detail depth (recommended correction values: down to -0.06 for SDXL).

  • Start/End-Percent Logic: Allows precise temporal control over when specific sampler settings influence the generation.

• LoRA Management:

  • Integrated LoRA Manager: Direct connection to local folder structures.

  • Trigger-Word Auto-Fill: Automatic import of tags (compatible with CivitAI metadata).

  • Multi-LoRA Support: Cascaded setup for multiple LoRAs without loss of quality.

Performance Optimization

• Any Switch & Bypasser Logic: Utilizing rgthree-nodes for quick deactivation of unused paths. Controlled via a global fast bypasser for groups or individual nodes.

• Hardware Efficiency: Optimized for systems ranging from 8GB VRAM (e.g., RTX 3070 Mobile) to high-end workstations. Render times for SDXL are approximately 3–5 seconds (depending on the system).

Installation & Requirements

• Base: ComfyUI (Portable or App version).

• Required Custom Nodes: * rgthree-comfy

  • ComfyUI-Impact-Pack

  • ComfyUI-Custom-Scripts (pysssss)

  • Portrait Master

• Models: Compatible with all SDXL 1.0 / Turbo / Lightning checkpoints.

Note: This workflow is modular. Extensions for Image-to-Image as well as connections to LLMs (Qwen/Florence2) can be toggled via the provided interfaces at any time.

DE: Dieser Workflow ist ein Geschenk an die Community. Ihr könnt ihn frei nutzen, modifizieren und eure Ergebnisse (Bilder) kommerziell verwenden. Ein Credit/Link zu unserem Kanal ist kein Muss, hilft uns aber enorm, weiterhin kostenlose Workflows für euch zu erstellen. Bitte verkauft diesen Workflow nicht 1:1 weiter – Keep it free for everyone!

EN: This workflow is a gift to the community. Feel free to use and modify it, and use your generated images for commercial purposes. While crediting our channel is not mandatory, it is highly appreciated and helps us keep creating free resources for you. Please do not resell this workflow as your own – keep it free for everyone!