My LoRAs 1.0
Role: GenAI Artist / Visual Creator
Tools: ComfyUI, Custom LoRA Training
Project Overview:
Achieving true character consistency is one of the biggest challenges in Generative AI. To push the boundaries of visual storytelling and art direction, I trained two custom LoRA models, "mmz guy" and "mmz lady", acting as my own original digital cast.
This project explores the intersection of machine learning and cinematic art direction. By building dedicated, multi-step ComfyUI pipelines, I established absolute control over these characters, ensuring their unique identities, complex gear, and hyper-realistic skin textures remain consistent across any generated scenario.
mmz_guy reference images:
mmz_lady reference images:
Technical Breakdown:
Custom LoRA Integration: Loaded specific custom-trained LoRA models tailored for distinct character archetypes: a rugged, atmospheric male lead ("mmz guy") and a precise, tactical female operative ("mmz lady").
Precision Conditioning & Sampling: Built a node graph with refined prompt conditioning and targeted sampling strategies to control moody lighting and cinematic composition.
High-Fidelity Detailing: Integrated the FaceDetailer node to maintain absolute anatomical consistency and high-resolution skin textures in close-up shots.
Production-Ready Upscaling: Engineered a final upscaling pass within the pipeline to ensure the output meets the high-resolution, high-fidelity standards required for professional visual content creation.