CivArchive
    Fuwa Wuffington | Anima | Personal Retrain - V2 · 3000 Steps
    Preview 146946449
    Preview 146946447
    Preview 146946452
    Preview 146946453
    Preview 146946454

    Fuwa Wuffington — Personal Retrain for Anima

    Fuwa belongs to GAi. This is my personally trained LoRA of his character.

    Fuwa is a pokesona based on a shiny Furfrou that GAi caught during a Pokémon X/Y stream. She went on to become one of his signature characters.

    WHY I TRAINED THIS VERSION

    I wanted to use Fuwa alongside other character and style LoRAs.

    In my setup, GAi's original LoRA tended to dominate the result and was difficult to combine with other LoRAs. I trained this version with the aim of keeping Fuwa recognizable while leaving more room for the rest of the image.

    That is the goal of this release. Results still depend on your checkpoint, prompt, LoRA strengths, and the other models you're combining.

    CHARACTER CREDIT

    Fuwa Wuffington is GAi's original character.

    Original creator and model:

    https://civarchive.com/models/2395240/gais-ocs-or-fuwa-wuffington

    Credit for the character and her design goes to GAi. My contribution is this particular LoRA training.

    AVAILABLE VERSIONS

    • V2 · 3000 Steps

    The earlier of the two checkpoints from my V2 training run.

    • V2 · 3600 Steps

    The same run after another 600 training steps.

    V2 refers to my own training revision, not a version of GAi's original release.

    Both checkpoints use the same trigger. Compare them with the same prompt and settings; the later checkpoint isn't automatically the better choice.

    TRIGGER WORD

    fuwanima

    Use this token for my retrain. GAi's original model has its own trigger words.

    Start with fuwanima, then describe the outfit, pose, expression, and setting you want.

    Example:

    fuwanima, solo, standing by a cafe window, holding a coffee cup, looking at viewer, soft afternoon light

    CHARACTER DETAILS

    Fuwa's design includes short black hair, red eyes, freckles, dark blue eyeshadow, bags under her eyes, and a tall, slender build.

    For her anthro appearance, describe the dog ears, tail, animal nose, and fur when needed. If you want the familiar outfit, add:

    white turtleneck sweater dress, black pantyhose, white flats

    Outfit and appearance tags are supporting descriptions, not additional trigger words. Start simple and reinforce whichever features need help.

    USING OTHER LORAS

    Combining Fuwa with other LoRAs is the reason I made this version.

    Try a strength of 1.0 as an initial setting, then adjust to suit your setup. Lower it if Fuwa starts taking over the style or passing her features to another character. Raise it if her identity isn't coming through clearly enough.

    For scenes with multiple characters, describe each character separately and keep their features clearly assigned. Some mixing can still happen; this retrain doesn't guarantee complete separation.

    EXAMPLE PROMPTS

    Simple character check:

    fuwanima, solo, full body, standing, relaxed posture, white turtleneck sweater dress, black pantyhose, white flats, simple background

    Quiet afternoon:

    fuwanima, solo, sitting by a cafe window, holding a coffee cup, thoughtful expression, warm afternoon light

    Different outfit:

    fuwanima, solo, oversized black jacket, jeans, walking along a rainy street at night, neon reflections

    COMPATIBILITY

    Trained for Anima. Different Anima checkpoints and combinations of LoRAs can produce different results.

    The aim is simple: recognizable Fuwa, with room for the rest of the prompt.

    Description

    The 3000-step checkpoint from my second Fuwa Wuffington training run.

    I trained this version to make Fuwa easier to use alongside other character and style LoRAs. In my setup, the original owner's LoRA had more influence than I wanted for those combinations.

    This is the earlier of the two published V2 checkpoints. Compatibility with other LoRAs is the training goal, rather than a guarantee for every combination.

    QUICK START

    Trigger: fuwanima

    Initial strength to try: 1.0

    Example:

    fuwanima, solo, full body, white turtleneck sweater dress, black pantyhose, white flats, relaxed posture, simple background

    Adjust the strength according to how clearly Fuwa comes through and how much room the other LoRAs need.

    VERSION DETAILS

    • Base model family: Anima

    • Training steps: 3000

    • Network rank: 6

    • Format: SafeTensors

    • Training software: AI Toolkit Experimental 0.13.23

    • File: Fuwa_Wuffington_v2_000003000.safetensors

    The 3600-step checkpoint continues the same training run for another 600 steps. Compare both using the same prompt and settings.

    V2 is my training revision. Fuwa Wuffington and her character design belong to GAi.

    Original creator's model:

    https://civitai.red/models/2395240/gais-ocs-or-fuwa-wuffington

    LORA
    Anima

    Details

    Downloads
    19
    Platform
    CivitAI
    Platform Status
    Available
    Created
    10/27/2026
    Updated
    11/14/2026
    Deleted
    -
    Trigger Words:
    fuwanima

    Files

    Fuwa_Wuffington_v2_000003000.safetensors