CivArchive
    Digitalized Arts Babe: AnnieXV LoRA - v1.0
    NSFW
    Preview 120712701
    Preview 120712706
    Preview 120712702
    Preview 120712698
    Preview 120712707
    Preview 120712695
    Preview 120712699
    Preview 120712700
    Preview 120712703
    Preview 120712704
    Preview 120712696
    Preview 120712697
    Preview 120712705

    Finally, a professional-grade Annie Leonhart that doesn't sacrifice flexibility for likeness! Trained by Digitalized Arts Studio, the DABAnnieXV LoRA is the result of 1,600 highly optimized steps on a curated dataset.

    Why download this?

    • Hyper-Realistic Eyes: Custom-trained iris patterns and pale blue eye accuracy.

    • 👗 Outfit Versatility: Perfect results for her civilian sweater/skirt combo and her Paradis military gear.

    • 🚀 High Rank: Trained at Dim 64 for maximum detail retention without the "deep-fried" look of lower-rank models.

    • 🎨 Style Compatible: Works beautifully with photorealistic and anime-style checkpoints alike.

    Trigger Keywords: a realistic photo of dabanniexv, blonde hair, pale blue eyes, short hair. Recommended Weight: 0.7 - 1.0

    Show us your best Annie renders in the gallery below! If you love the model, please leave a 5-star review to help the community find it!

    License & Terms of Use

    Usage Restriction: This model is strictly for personal use only.

    +1

    Prohibitions: Commercial use, selling generated images for profit, and redistribution or re-uploading of this model to other platforms is prohibited.

    +1

    Dataset Access: If you are interested in purchasing the high-quality dataset used to train this LoRA, it is available for direct purchase on our official website.

    +1

    Official Website: www.digitalizedarts.com

    Description

    The training configuration for DABAnnieXV is built for high-detail character consistency on the Stable Diffusion v1.5 base.

    • Model Architecture: A "Standard" LoRA using the networks.lora module with a Network Dimension (Rank) of 64 and Alpha of 32 for balanced learning.

    • Training Specs: The model was trained for 1,600 steps at a learning rate of 0.0001, utilizing the AdamW8bit optimizer and cosine lr_scheduler for smooth convergence.

    • Resolution & Precision: Optimized for a native resolution of 1024x1536 (likely using aspect ratio buckets from 512 to 2048) and trained using bf16 mixed precision.

    • Dataset Characteristics:

      • Captions: Uses .txt extensions with shuffled captions to improve flexibility.

      • Key Trigger/Tags: Focuses on "dabanniexv," blonde hair, pale blue eyes, detailed iris patterns, and specific outfits like sweaters and blue skirts.

      • Settings: Includes gradient checkpointing and latent caching to optimize VRAM usage during the training process.

    LORA
    SD 1.5

    Details

    Downloads
    47
    Platform
    CivitAI
    Platform Status
    Available
    Created
    2/11/2026
    Updated
    4/27/2026
    Deleted
    -
    Trigger Words:
    dabanniexv, woman, blonde hair, pale blue eyes, looking at viewer

    Files

    Final - DABAnnieSD1.52 - W.out.zip

    Mirrors

    Available On (1 platform)

    Same model published on other platforms. May have additional downloads or version variants.