CivArchive
    Mocha-tan (Mwocha / Mocha Kaneko) | Anima - 3000 Steps
    NSFW
    Preview 148716820
    Preview 148716819
    Preview 148716825
    Preview 148716828
    Preview 148716834

    Mocha-tan — Character LoRA for Anima

    The coffee cat herself: Mocha-tan, also known as Mwocha and Mocha Kaneko.

    This is my personally trained character LoRA for Anima, made to bring Mocha into different outfits, settings, compositions, and illustration styles.

    The character and original VTuber model are credited below. My contribution is the LoRA training. This is an unofficial release and does not imply affiliation with or endorsement by Mocha or her model creators.

    CHARACTER & ORIGINAL MODEL CREDITS

    Mocha-tan / Mwocha / Mocha Kaneko

    VTuber: @mwochaaa

    Official website:

    https://mwocha.carrd.co/

    YouTube:

    https://www.youtube.com/@mwochaaa

    Twitch:

    https://www.twitch.tv/mwocha

    X / Twitter:

    https://x.com/mwochaaa

    Original VTuber model illustration / art mama:

    @ayaoshiro

    https://x.com/ayaoshiro

    Original Live2D rigging / rig papa:

    @pat__cha

    https://x.com/pat__cha

    Please visit their original accounts and support their work.

    THE CHARACTER

    Features to look for include:

    • Long brown hair with contrasting white streaks

    • Red eyes

    • Brown cat ears with light inner tufts

    • A distinctive looped ahoge

    • Paired hair clips resembling whiskers

    • Small facial moles beneath the eyes

    • A large fluffy brown tail with a white tip

    • A short, compact silhouette

    Mocha is a catgirl with human skin, hands, and feet. If a checkpoint starts turning her into a fully fur-covered character, reinforce “cat girl” rather than “anthro.”

    Her identity is the focus of the LoRA; her clothes are not fixed. Describe the outfit, expression, pose, and surroundings you want.

    AVAILABLE CHECKPOINTS

    3000 Steps

    The earlier release from this training run.

    3600 Steps

    A second checkpoint from the same run, with 600 additional training steps.

    Both use the same activation word: mochanima.

    Try both with matching prompts, seeds, and generation settings. Choose whichever gives you the balance of character resemblance and stylistic flexibility you prefer.

    GETTING STARTED

    Trigger:

    mochanima

    Initial LoRA strength:

    1.0

    Basic character check:

    mochanima, 1girl, solo, standing, white blouse, blue skirt, looking at viewer, smile, simple background. mochanima stands facing the viewer in a white blouse and blue skirt, smiling against a plain background.

    Begin with the activation word and your scene. Add supporting appearance tags only when a feature needs reinforcement.

    Optional supporting tags:

    cat girl, cat ears, brown hair, streaked hair, red eyes, ahoge, hairclip, mole under eye, tail

    These describe the character; they are not additional activation words.

    CLOTHES, STYLES & SETTINGS

    The training references include different clothing, poses, compositions, and scenes with companions. You can prompt casual clothes, formal outfits, fantasy clothing, or other wardrobes instead of relying on one default costume.

    Artist tags can be used to explore different interpretations. Keep the character prompt consistent when comparing artists so you can see how much the rendering changes.

    EXAMPLE PROMPTS

    A quiet afternoon:

    mochanima, 1girl, solo, sitting, reading, holding book, green cardigan, white blouse, blue skirt, black pantyhose, loafers, library, window, sunlight. mochanima reads an open book on a library window seat, with afternoon light falling across her cardigan and the pages.

    Rainy evening:

    mochanima, 1girl, solo, walking, looking back, red jacket, black pants, black boots, holding umbrella, transparent umbrella, rain, street, neon lights, night. mochanima looks back while walking beneath a transparent umbrella, surrounded by colorful reflections on a rainy city street.

    Festival night:

    mochanima, 1girl, solo, smile, red kimono, floral print, obi, wide sleeves, geta, holding fan, paper lantern, festival, fireworks, night. mochanima smiles beneath glowing festival lanterns, holding an open fan beside her shoulder as fireworks spread across the sky.

    STRENGTH & CONSISTENCY

    The showcase images use a LoRA weight of 1.0.

    Reduce the weight if the LoRA overwhelms your chosen style or carries Mocha’s features into another subject. If resemblance is weak, first check the trigger and reinforce the specific missing feature.

    For scenes with multiple people, describe each subject separately. Assign Mocha’s outfit, position, and actions clearly, and check that her ears, tail, or hairstyle do not transfer to her companion.

    COMPATIBILITY

    Anima-family character LoRA.

    The supplied showcase images were generated with Nova Furry AM v5.5.2-9B. Results will vary with the checkpoint, prompt, artist tags, LoRA strength, and other resources in your workflow.

    Different clothes, different surroundings, same coffee cat.

    Description

    The 3000-step Mocha-tan checkpoint for Anima.

    This is the earlier of the two releases from the same training run. The five showcase images prepared for this version use this checkpoint.

    QUICK START

    Activation word: mochanima

    Starting strength: 1.0

    Example:

    mochanima, 1girl, solo, green cardigan, white blouse, blue skirt, sitting, holding book, library, sunlight.

    mochanima sits beside a library window with an open book in her hands, illuminated by soft afternoon sunlight.

    TRAINING DETAILS

    • Training base: circlestone-labs/Anima-Base-v1.0-Diffusers

    • Saved training step: 3000

    • Epochs: 25

    • LoRA rank / alpha: 32 / 32

    • Learning rate: 2e-5

    • Training resolution: 1024 with aspect-ratio buckets

    • Character dataset: 60 images

    • Regularization dataset: 200 images

    • Character masks used during training

    • Precision: BF16

    • Training software: AI Toolkit Experimental 0.13.23

    The step count includes the regularization updates.

    SHOWCASE SETTINGS

    • Checkpoint: Nova Furry AM v5.5.2-9B

    • LoRA strength: 1.0

    • Sampling steps: 30

    • CFG: 4

    • Sampler: er_sde

    • Scheduler: sgm_uniform

    • Upscaler: RealESRGAN_x4plus, 4×

    These document the showcase workflow rather than a requirement for using the LoRA.

    CREDITS

    VTuber / character:

    Mocha-tan — @mwochaaa

    https://mwocha.carrd.co/

    https://www.youtube.com/@mwochaaa

    https://www.twitch.tv/mwocha

    https://x.com/mwochaaa

    Original VTuber model illustration:

    @ayaoshiro

    https://x.com/ayaoshiro

    Original Live2D rigging:

    @pat__cha

    https://x.com/pat__cha

    LoRA training by me

    LORA
    Anima

    Details

    Downloads
    26
    Platform
    CivitAI
    Platform Status
    Available
    Created
    11/12/2026
    Updated
    11/14/2026
    Deleted
    -
    Trigger Words:
    mochanima

    Files

    MochaVT_v1_000003000.safetensors

    Mirrors