MFMXL is an SDXL LoRA designed to generate realistic MFM (Male/Female/Male) threesome compositions. It assists in creating consistent three-person positioning across multiple scenarios while remaining flexible enough to work with a variety of SDXL checkpoints.This model has been extensively tested across hundreds of generations and tuned for realistic compositions. As with any complex multi-character LoRA, results depend on checkpoint, LoRA strength, and prompting. Recommended settings and testing notes are included below. ------------------------------------------------------------------------------------------------------- Recommended Settings<h2 id="base-models">Base Models</h2>Recommended:<ul><li>Pornmaster</li><li>Lustify</li><li>Juggernaut XL</li></ul><h2 id="lora-strength">LoRA Strength</h2>Recommended 0.5 to 0.8<h2 id="generation-settings">Generation Settings</h2>Recommended starting settings:<ul><li>CFG: 5</li><li>Steps: 35</li></ul><h1 id="supported-concepts">Supported Concepts</h1>Current positions include:<ul><li>Blowjob</li><li>Cowgirl</li><li>Missionary</li><li>Doggystyle</li><li>Side-Lying</li><li>Cumshot</li><li>POV (limited)</li></ul><h1 id="aspect-ratios">Aspect Ratios</h1>The training dataset was primarily built from 3:2 and 2:3 images with a smaller percentage of additional aspect ratios.Testing showed surprisingly strong performance at 1:1, while landscape generations remain the most challenging.<h1 id="prompting">Prompting</h1>This LoRA responds best to simple prompts. Tags: MFM, Blowjob, Cowgirl, Cumshot, Doggystyle, Missionary, POV, Side-lying

MFMXL  (MFM Threesomes)

Initial public release of MFMXL. Trained for realistic MFM threesome compositions across multiple positions with an emphasis on flexible prompting and composition consistency. See the model description for recommended settings and known limitations.

Files with SHA256: 84d33dd138dffced3fcdb8bbc8df10955eac1847d15a3b9c14caf46e7ecefdf4