ArtificialGuyBR

Home / Blog / AI video

Yarnwoven Style LoRA Transforms Videos into Stop-Motion Yarn Worlds

LoRA for Wan 2.2 T2V-14B that converts scenes into yarn-textured stop-motion animations.

2 sources cited AI video

On this page

  1. What It Does
  2. Model Details
  3. Setup for ComfyUI
  4. Example Prompts
  5. Using with Diffusers
  6. Common Errors and Fixes
  7. FAQ
  8. Sources

Yarnwoven Style LoRA for Wan Video Generation

TL;DR

What It Does

This LoRA adapter applies a distinctive yarnwoven texture to all elements in your video prompts. Instead of photorealistic or painterly renders, the model produces scenes where characters, objects, and environments appear as if handcrafted from colored yarn. The result resembles traditional stop-motion animation using felt or fabric materials.

Yarnwoven Style sample - ninja scene

Yarnwoven Style sample - mechanic scene

Model Details

Start prompts with the trigger phrase followed by your scene description for best results. The model interprets any subject matter through a yarn-textured lens, adding stop-motion animation characteristics.

Setup for ComfyUI

  1. Download [WAN2]YarnwovenStyle-Redmond_high_noise.safetensors
  2. Place in ComfyUI/models/loras/
  3. Load your base Wan model
  4. Add a Load LoRA node after model loader
  5. Connect MODEL output to KSampler
  6. Set strength_model to 0.8-1.0

Example Prompts

Yarnwoven Style. In a world made of yarn, a ninja made of yarn throws shuriken made of yarn, leaving trails made of yarn through a bamboo forest made of yarn. Stop motion animation.
Yarnwoven Style. In a world made of yarn, a mechanic made of yarn repairs an engine made of yarn with tools made of yarn in a garage made of yarn. Stop motion animation.
Yarnwoven Style. In a world made of yarn, a Viking made of yarn rows a longship made of yarn through stormy seas made of yarn, lightning made of yarn striking. Stop motion animation.

Yarnwoven Style sample - Viking scene

Using with Diffusers

from diffusers import WanPipeline
from peft import PeftModel

pipe = WanPipeline.from_pretrained("Wan-AI/Wan2.2-T2V-A14B")
pipe.load_lora_weights("artificialguybr/YarnwovenStyle-Redmond-WAN2-T2V-14B")

Common Errors and Fixes

Error | Cause | Fix | Source

|---|---|---|---

Black/blank output | LoRA not connected to KSampler | Connect LoRA MODEL output to KSampler model input | HF card No style change | Missing trigger phrase | Start prompt with Yarnwoven Style. exactly | HF card Over-cooked texture | Strength too high | Lower strength_model to 0.8-0.9 | HF card

FAQ

Q: What base model is required? A: Wan-AI/Wan2.2-T2V-A14B (14B parameter text-to-video model).

Q: Does it work with Wan 2.1 or other Wan variants? A: Optimized for Wan 2.2 T2V-A14B; results may vary on other versions.

Q: Can I use it for text-to-image? A: Tagged for text-to-image in HF but trained for video; image results untested.

Q: What license applies? A: CC0-1.0 (public domain) — free for any use including commercial.

Sources