[โ๐ฆ๐ค๐ฅ๐๐๐ฃ๐ฎ| ๐ฝ๐ ๐ฃ๐ฅ๐๐๐ฅ๐] - ๐จ๐๐๐๐ฆ๐ค๐ฅ๐ฃ๐๐ ๐ฆ๐ค ๐๐/๐ฆโ๐ ๐๐ช ๐๐/๐ผ๏ธ๐ธ๐๐๐๐ ๐๐ โ๐ธ
This is a high-fidelity LoRA designed for Illustrious-XL/Pony XL/Anima models. It was trained on a curated synthetic dataset to ensure maximum flexibility and character consistency while maintaining the high-aesthetic standards of the base model.
โผ๏ธIMPORTANTโผ๏ธ
โ-[OUTDATED VERSION]
๐จ๐ฆ๐ผ๏ธ-[RECOMMEND TO USE]
๐ฎ๐๐๐น๐๐ ๐ช๐๐๐๐๐พ๐๐: ๐ฏ๐ฝ๐ ๐ ๐๐๐๐๐๐ ๐๐๐๐๐ฝ๐๐๐พ๐ธ
This LoRA is designed to capture the specific "Rough Rider" set look. It excels at rendering her signature desert-ready gear
๐จ๐ฆ๐๐ ๐ฑ๐พ๐๐๐ถ๐ ๐น๐๐ถ๐๐๐๐๐
The Outfit: Distressed denim jeans, a brown leather utility belt with pouches, and a cropped flannel or tactical shirt.
Key Accessories: Her iconic wide-brimmed cowboy hat and sturdy leather boots.
The Vibe: A blend of tactical survivor and classic Western outlaw.
๐ ๐๐ค๐๐๐ โ๐๐๐ ๐๐๐๐๐๐๐ฅ๐๐ ๐๐ค
โ๏ธBase Model: ๐จIllustrious
Trigger Word:
RustlerFNLoRA Weight:
0.7 - 0.8Clip Skip:
2Recommend to use: Nova 3DCG XL, WAI-illustrious-SDXL
โ๏ธ Base Model: ๐ฆPony XL
Trigger Word:
RustlerFNRecommended Weight:
0.8 - 0.9Recommend to use: WAI-ANI-PONYXL, Perfect Pony XL
โ๏ธ Base Model: ๐ผ๏ธAnima
Trigger Word:
RustlerFNRecommended Weight:
0.8 - 0.9Recommend to use: NOVA ANIME AM, WAI-ANIMA
๐ ๏ธ โ๐ฃ๐ ๐๐ก๐ฅ๐๐๐ ๐พ๐ฆ๐๐๐
This model is highly responsive to descriptive tags. For the best results, use a mix of the trigger word and specific outfit/pose descriptions.
Main Trigger:
RustlerFNCore Look:
blonde hair, green eyes, long hair, lipsSignature Outfit:
cowboy hat, plaid shirt, tied shirt, sleeves rolled up, midriff, navel, denim shorts, blue shorts, beltTactical Gear:
holster, gun, gloves, knee pads, elbow pads, thigh strap
Negative Tags:
low quality, blurry, deformed, text, watermark
๐ ๐๐๐๐๐๐๐๐๐ ๐ป๐๐ฅ๐๐๐๐ค
Dataset: Created using a high-quality synthetic workflow to generate diverse angles and lighting from a single reference source.
Training: Optimized using
bfloat16for precision.Train Model: WAI-illustrious-SDXL v.17/ WAI-ANI-PONYXL v.14/Anima-Base-v1.0
๐ฌ โ๐ ๐ฅ๐๐ค & ๐ฝ๐๐๐๐๐๐๐
I am constantly refining the dataset generation process. If you enjoy the model, please post your generations and leave a review-it helps me understand how to improve future versions!
If you run into issues with character bleeding or specific clothing items, let me know in the discussion.
Description
Version 1.0 Update: Massive Quality & Flexibility Overhaul!
I'm incredibly excited to release this major update! For this version, I completely went back to the drawing board to rebuild the dataset from scratch and optimized the training settings. The result is a much smarter, more obedient, and higher-quality LoRA.
๐๏ธ The Dataset Rebuild
Completely Redone Dataset: I scrapped the old images and curated a brand new, strictly high-quality dataset.
Pristine Tagging: Completely re-captioned the data, fixing tag bleed and removing messy, conflicting concepts. The model now actually understands exactly what you are asking for.
Sharper Source Images: Upgraded and perfectly cropped the source materials to capture tiny details without the blur or artifacting from v0.1
โ๏ธ Optimized Training Settings
Refined Hyperparameters: Tweaked the learning rate, network rank (dim), and alpha settings to prevent the LoRA from getting "fried" or overbaked.
Better Epoch Selection: Tested multiple epochs to find the exact sweet spot where the concept is perfectly learned without destroying the base model's flexibility.
โจ The Results: Quality & Flexibility
Stunning Output Quality: Noticeably sharper details, better textures, and way fewer visual artifacts compared to the previous version.
Ultimate Prompt Flexibility: The LoRA is no longer "stubborn"! You can now easily change outfits, poses, lighting, and environments without the model fighting back or defaulting to the training data.
Plays Nice with Others: Blends seamlessly with other LoRAs and a massive variety of different base checkpoints.