skills.rest
artokun
artokunCOMMUNITY
Explore all skills by this author →
Website𝕏@artokunGitHub

anima-lora-trainer

Train custom anime LoRA adapters on the ANIMA base model using kohya sd-scripts.

715|111|Updated Feb 15, 2026
Search Tags:#fine-tuning#training#anime#Anima#ComfyUI#Lora#kohya sd-scripts
One-click install
npx skills add https://github.com/artokun/comfyui-mcp --skill anima-lora-trainer
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: anima-lora-trainer
Source: https://github.com/artokun/comfyui-mcp/tree/main/plugin/skills/anima-lora-trainer
Command: npx skills add https://github.com/artokun/comfyui-mcp --skill anima-lora-trainer

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Training a custom character or style LoRA for the ANIMA anime diffusion model requires correctly configuring kohya sd-scripts, dataset TOMLs, VRAM-constrained parameters, and model paths, which is error-prone to do by hand.

Core Features & Use Cases

  • Guided Trainer Setup: Walks through Citron's Gradio-based Anima LoRA Trainer installation on Windows or RunPod, including GPU-specific PyTorch CUDA wheel selection and model downloads.
  • Dataset & Parameter Configuration: Explains the flat-folder image-plus-caption dataset format and provides default training parameters (network_dim 32, resolution 768, AdamW8bit, bf16) that fit in roughly 6GB of VRAM.
  • ComfyUI Integration: Shows how to load the resulting .safetensors LoRA into the anima-base ComfyUI workflow via LoraLoaderModelOnly.
  • Use Case: An artist wants a LoRA of their original character for ANIMA image generation; this Skill guides them from dataset captioning through training to using the LoRA in ComfyUI.

Quick Start

Help me set up Citron's Anima LoRA Trainer and train a character LoRA from my folder of captioned images.

Frequently Asked Questions about anima-lora-trainer

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I train a LoRA for the ANIMA anime model?▼

Use Citron's Anima LoRA Trainer, a Gradio UI wrapping kohya sd-scripts. Prepare a flat folder of images with matching .txt captions, configure parameters like network_dim 32 and resolution 768, then start training to produce a .safetensors LoRA.

How much VRAM is needed to train an Anima LoRA?▼

Default settings train on roughly 6GB of VRAM using network_dim 32, resolution 768, batch size 1, gradient checkpointing, and latent caching. On out-of-memory errors, lower network_dim to 8 or resolution to 512.

What dataset format does kohya sd-scripts LoRA training require?▼

A flat folder of images (.jpg, .png, .webp, and similar) where each image has a matching .txt caption file with the same basename. Captions use Danbooru-style tags or natural language, with caption_extension set to .txt in the dataset TOML.

Can I run the Anima LoRA trainer on RunPod or Linux?▼

Yes, the CITRON_ANIMA_LORA_TRAINER-RUNPOD-V2.sh script installs everything into /workspace on RunPod and patches the Gradio server to 0.0.0.0. Expose HTTP port 7860 to access the training UI through the RunPod proxy.

How do I use a trained Anima LoRA in ComfyUI?▼

Copy the trained .safetensors file into ComfyUI's models/loras directory and load it in the anima-base workflow with a LoraLoaderModelOnly node. Typical strength is 0.7 to 1.0, and it can be stacked with the turbo LoRA for fast 12-step generation.

Related Skills

Explore More Design & Creative Skills→
fabioc-alohafabioc-aloha
1

flux-brand-finetune

Train custom FLUX LoRA fine-tunes on Replicate from curated image datasets.

Community
Advanced
erfianugraherfianugrah
1

lora-train

Automate SDXL or Flux LoRA training, captioning, and ComfyUI deployment via HTTP proxy.

Community
Intermediate
redbananastudiosredbananastudios

lora-evaluate

Generate a structured test grid and score LoRA image outputs for consistency, fidelity, overfitting, and artefacts.

Community
Advanced
cuba6112cuba6112

unsloth-lora

Configure LoRA and rsLoRA parameters for efficient 16-bit LLM fine-tuning.

Community
Intermediate