Portrait Master for Flux1 – High-Fidelity Face Workflow Suite 1.0
Portrait Master Flux1 is a modular workflow suite for ComfyUI, developed for the production of photorealistic, character-consistent portraits at high resolution. Built around the Flux1 API and enhanced with semantic inpainting, expression control, and layered refinement, the system is ideal for professional creators in:
- Digital branding and synthetic influencer development
- AI-generated fashion, editorial, and commercial visual assets
- Sports and lifestyle video production using virtual talents
- Dataset creation and LoRA fine-tuning pipelines
The suite includes three specialized workflows, a Redux reference template, and training assets designed to scale from prompt to dataset to deployable LoRA.
Who It Is For
- Content creators building AI influencers and virtual brand models
- Studios producing AI-based video ads with virtual athletes or presenters
- Teams generating fashion/e-commerce visuals at scale
- Developers and researchers preparing LoRA training datasets
- Any creator seeking control, resolution, and repeatability in AI face generation
Workflow Versions Included
The suite includes three distinct ComfyUI workflows, each optimized for a specific level of control and production use.
1. PortraitMaster_Flux1_Expression.json
Complete workflow with expression control via Expression Editor (PHM)
Includes all modules: prompt injection, Flux1 guidance, latent noise injection, semantic masking (eyes, chin, lips), facial expression control from reference, dual upscaling to 6000px+, and full image grading (LUT, grain, exposure).
Ideal for creators needing expressive identity variation (smile, wink, blink, phoneme control) and fine-grain editing.
2. PortraitMaster_Flux1_Core.json
Full workflow without Expression Editor
Same features as above, excluding the expression transfer module. Preserves all inpainting, skin detailing, and post-production stages.
Recommended for clean neutral faces, consistent character identity, or when working with fixed LoRA expressions.
3. PortraitMaster_Flux1_Lite.json
Streamlined generation pipeline
Includes only the essential stages: prompt + Flux1 conditioning, latent refinement, and single upscale.
Perfect for rapid drafts, low-VRAM systems, or when expression/post-processing is not needed.
Additional Tools Included from (Generative Workflow Suite)
PortraitMaster_BatchGen.json
Workflow for batch-generating portrait variations using a reference image and a .txt file with prompts.
Compatible with GPT-generated instructions from:
Character Prompt Catalogue GPT
https://chatgpt.com/g/g-67f099ab7b64819187c8f0c3312ff3bb-character-prompt-catalogue
Redux_Template.psd
Photoshop base file to prepare Redux-compatible reference images. Includes layered regions and two example characters.
LoRA Training Assets
-
LoRA_Training_Guide_KohyaSS.txt– Step-by-step LoRA training workflow using Kohya_ss, including folder organization, repeat strategy, and optimizer recommendations. -
LoRA_config.json– Kohya_ss config for frontal image training
These resources support a complete two-stage training process:
- Create the base LoRA for frontal consistency
- Expand into poses and expressions for secondary LoRA generalization
System Requirements
- Python: 3.10 or higher
- PyTorch: 2.0.1 or higher
- ComfyUI: Latest master version
- GPU: NVIDIA RTX 3090 / 4090 or equivalent (24GB VRAM or higher recommended)
- Flux1 API: Access required; VPN to the United States may be necessary