Sign In

[VRChat] Karin | カリン | あまとうさぎ

0

Updated: Sep 9, 2026

charactervrchatkarin

Download

1 variant available

bf16 SafeTensor

karin-nonchibi-final.safetensors

BF16, good balance • 104.04 MB

Verified:

Type
LoRA
Stats

71

0

Reviews
Published

Sep 9, 2026

Base Model

Anima

Hash
AutoV2
EB1E78DBAA
Trigger Words
karin \(vrchat\)
default creator card background decoration
Followers - 7

7

Likes - 42

42

License:

Anima

The Anima Model is licensed by CircleStone Labs LLC. Copyright CircleStone Labs LLC. IN NO EVENT SHALL CIRCLESTONE LABS LLC BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH USE OF THIS MODEL.

Built on NVIDIA Cosmos

ComfyUI_temp_aypxk_00007__cleaned.png

Introduction

This LoRA is trained on Karin (カリン), a VRChat avatar created by Amatousagi (あまとうさぎ).

Usage

Recommended LoRA strength: 1.0
Trigger word: karin (vrchat)

In ComfyUI, escape the parentheses:

karin \(vrchat\)

Default outfit:

white sweatshirt, long sleeves, pink sleeve bows, black suspenders, black pleated skirt, white sneakers, fox tail

Training Details

  • Trainer: sorryhyun/anima_lora v1.17.1

  • Dataset: 51 captioned images, with 23 higher-quality images sampled twice and 28 images sampled once

  • LoRA rank / alpha: 32 / 128

  • Learning rate: 2e-5

  • Optimizer: AdamW

  • Scheduler: Cosine, with 5% warmup

  • Batch size: 1

  • Released checkpoint: Step 1,392

  • Precision: BF16

  • Resolution: Aspect-ratio buckets at approximately 1 megapixel

  • GPU: NVIDIA RTX 3090

简介

这是基于 Anima Base 1.0 训练的 Karin(カリン)角色 LoRA。Karin 是由 あまとうさぎ 制作的 VRChat 虚拟形象。

使用

推荐 LoRA 权重: 1.0
激活词: karin (vrchat)

在 ComfyUI 中需要转义括号:

karin \(vrchat\)

常服:

white sweatshirt, long sleeves, pink sleeve bows, black suspenders, black pleated skirt, white sneakers, fox tail

训练参数

  • 训练工具: sorryhyun/anima_lora v1.17.1

  • 数据集: 51 张带 caption 的图片,其中 23 张高质量图片重复采样 2 次,其余 28 张采样 1 次

  • LoRA rank / alpha: 32 / 128

  • 学习率: 2e-5

  • 优化器: AdamW

  • 学习率调度: Cosine,5% warmup

  • Batch size: 1

  • 发布权重:1,392

  • 训练精度: BF16

  • 训练分辨率: 约 1 M 像素的宽高比分桶

  • 显卡: NVIDIA RTX 3090