Description:This LoRA is dedicated to recreating the vibrant and heroic Star Guardian skin for Poppy from League of Legends.This is a UNet-only LoRA (<code>network_train_unet_only = true</code>), meaning it was trained exclusively on the visual aspects of the character's design. It does not teach the model a new "concept" of Poppy but instead excels at applying her Star Guardian outfit, hairstyle, and signature hammer over a well-prompted character.For optimal compatibility, this LoRA was trained on NoobAI XL Vpred 1.0 and works best with v-prediction models.<hr /><h2 id="usage-instructions">Usage Instructions</h2><h3 id="activation-and-prompting">Activation &amp; Prompting</h3>Since the text encoder was not trained, you must describe the character in your prompt. Use the primary trigger word, then add descriptive tags for her appearance and outfit.<ul><li>Primary Trigger: <code>sgpoppy</code></li><li>Essential Character Tags: <code>yordle</code>, <code>blue skin</code>, <code>purple hair</code>, <code>twintails</code>, <code>blue eyes</code>, <code>pointy ears</code>, <code>fang</code></li><li>Essential Outfit Tags: <code>star guardian uniform</code>, <code>armor</code>, <code>breastplate</code>, <code>skirt</code>, <code>blue gloves</code>, <code>pantyhose</code>, <code>white legwear</code>, <code>kneeboots</code></li><li>Weapon Tag: <code>hammer</code></li></ul><h3 id="important:-lora-weight">Important: LoRA Weight</h3>This LoRA was trained with an alpha (<code>64</code>) that is double its dimension (<code>32</code>). This makes the LoRA very strong.<ul><li>Works plenty fine with a weight of <code>1.0</code>.<ul><li>Example: <code>&lt;lora:SGPoppy:1.0&gt;</code></li></ul></li></ul><h3 id="recommended-settings">Recommended Settings</h3><ul><li>Model: Any SDXL v-prediction model (e.g., NoobAI-XL, Illustrious-XL).</li><li>Sampler: Euler a CFG++</li><li>Steps: 20</li><li>CFG Scale: 1.4-1.7</li></ul><hr /><h2 id="outfit-and-style-description">Outfit &amp; Style Description</h2>This LoRA accurately reproduces the key elements of Poppy's Star Guardian design:<ul><li>The Outfit: Captures her distinct blue, white, and gold magical girl uniform, complete with the star emblem, large bows, and armored gauntlets/boots.</li><li>Character Features: Reliably generates her yordle physique, vibrant blue skin, and signature purple hair styled in large twintails.</li><li>The Starlight Hammer: Excellently reproduces her iconic, oversized golden hammer, maintaining its grand scale in proportion to her small stature.</li></ul>The style is sharp and illustrative, perfect for creating high-impact character art that reflects Poppy's determined and heroic personality.<hr /><h2 id="training-details">Training Details</h2>This LoRA was trained on a focused dataset of 29 images with the following notable parameters:<ul><li>Base Model: <code>NoobAI XL Vpred 1.0</code></li><li>Resolution: 1024x1024 (with bucketing enabled)</li><li>Network Rank (Dimension): 32</li><li>Network Alpha: 64<ul><li>Note: A 2:1 alpha-to-dim ratio was used, amplifying the LoRA's trained effect. This is why lower weights are recommended during inference.</li></ul></li><li>Training Target: UNet Only</li><li>Optimizer: AdamW8bit</li><li>Learning Rate: <code>4e-4</code> (4×10−4)</li><li>LR Scheduler: <code>cosine_with_restarts</code> (5 cycles)</li><li>Advanced Noise Scheduling: An extremely high <code>min_snr_gamma = 80</code> was used alongside <code>zero_terminal_snr = true</code>, forcing the model to focus intensely on high-level composition and detail.</li><li>Max Training Epochs: 50</li></ul>

Star Guardian Poppy

Version 1.0 - Initial ReleaseThis is the first public version of the Star Guardian Poppy Style/Outfit LoRA.<ul><li>Training Base: The model was trained on a focused dataset of 29 images using NoobAI XL Vpred 1.0 as the base model.</li><li>Training Type: This is a UNet-Only LoRA. It is designed to replicate the visuals (outfit, hair, hammer) of the character and relies on your prompt to create the underlying anatomy and features.</li><li>Key Characteristics:<ul><li>The LoRA was trained with a 2:1 alpha-to-dimension ratio (<code>α=64</code>, <code>dim=32</code>), making it very potent. It is crucial to use lower weights (e.g., 0.6-0.8) to avoid oversaturation.</li><li>Advanced training techniques, including a very high <code>min_snr_gamma</code> of 80, were used to maximize detail and adherence to the source style.</li></ul></li></ul>The goal of this initial version is to provide a strong, high-quality LoRA for generating the full Star Guardian Poppy outfit with excellent accuracy.

Files with SHA256: 3daad93d815dcb057344c30091011b6d02b4338fb379e2b81718a89a0a42ccbc