[πΌπππ₯π β€ππππβοΈ| π½π π£π₯πππ₯π] - π¨ππππ¦π€π₯π£ππ π¦π€ ππ/π¦βπ ππͺ ππ ππ βπΈ/πΌοΈπΈππππ ππ βπΈ
This is a high-fidelity LoRA designed for Illustrious-XL/Anima 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.
βΌοΈIMPORTANTβΌοΈ
β-[OUTDATED VERSION]
π¨π¦πΌοΈ-[RECOMMEND TO USE]
βοΈπππΉππ πͺπππππΎππ: πΈππΎππ π΅πΆπΉπΎπ (πΉπππππΎππ)
This LoRA (Low-Rank Adaptation) model captures the distinct aesthetic of the Elite Zadie skin from Fortnite. Originally introduced as an exclusive competitive tournament reward (via the Elite Zadie Reload Cup) and later added to the Item Shop, this character design features a tactical, stealth-oriented design paired with a stark, modern sci-fi look.
The model is highly effective for rendering character concepts, concept art, and detailed portraits maintaining consistent clothing patterns and facial structure.
π¨π¦ππ π±πΎπππΆπ πΉππΆπππππ
Hair: Long, flowing stark-white or silver hair, typically styled straight with a clean center part.
Face: Sharp, symmetric facial features with intense, light-colored eyes (often blue or light gray) and well-defined makeup.
Attire: A matte-black, form-fitting tactical bodysuit/catsuit. The suit features subtle segmented armor plates on the chest, clean stitching lines, a high-collared neck, and utility accents such as buckles, straps, and an utility belt.
π ππ€πππ βπππ πππππππ₯ππ ππ€
βοΈπ΅πΆππ πππΉππ: π¨πΌπππππππΎπππ
Trigger Word:
EliteZadieFNLoRA Weight:
0.7 - 0.8(Start at0.7for maximum accuracy)Clip Skip:
2Recommend to use: Nova 3DCG XL, WAI-illustrious-SDXL
βοΈπ΅πΆππ πππΉππ: π¦π«πππ π³πΏ
Trigger Word:
EliteZadieFNRecommended Weight:
0.8 - 0.9Recommend to use: WAI-ANI-PONYXL, WAI-SemiReal, BlenderMix
βοΈ Base Model: πΌοΈAnima
Trigger Word:
EliteZadieFNRecommended Weight:
0.8 - 0.9Recommend to use: NOVA ANIME AM, WAI-ANIMA
π οΈ βπ£π ππ‘π₯πππ πΎπ¦πππ
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:
EliteZadieFNCore Look:
1girl,solo,long hair,white hair,grey eyes,lips,looking at viewerSignature Outfit:
black bodysuit,tactical clothes,turtleneck,bodysuit,tight clothesLegwear & Shoes:
leggings,thighsDetails & Accessories:
mask on belt,belt,buckle,hip mask,glowing eyesStyle:
3d
Negative Tags: low quality, blurry, deformed, text, watermark
π πππππππππ π»ππ₯ππππ€
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
π¬ βπ π₯ππ€ & π½πππππππ
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.
ποΈ 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
βοΈ 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.
β¨ 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 an
FAQ
Comments (1)
Hush - Fortnite please