ArtificialGuyBR

Home / Blog / Training LoRAs

Crochet Amigurumi LoRA for FLUX.2 Klein

Generate consistent crochet plush toys with FLUX.2 Klein using this LoRA. Trigger words, settings, and examples.

2 sources cited Training LoRAs

On this page

  1. What This Model Does
  2. Training Details
  3. How to Use
  4. Recommended Settings
  5. Common Errors and Fixes
  6. FAQ
  7. Sources

Crochet Amigurumi LoRA for FLUX.2 Klein

What This Model Does

This LoRA teaches FLUX.2 Klein to generate crochet amigurumi — the Japanese art of knitting or crocheting small, stuffed yarn creatures. The model produces soft, plush toy characters (bunnies, bears, custom creatures) with visible yarn texture and stitching detail. The style is consistent: rounded forms, soft color palettes (cream, beige, tan), and the distinctive look of hand-crocheted items.

Crochet amigurumi bunny with floppy ears and detailed yarn texture

Training Details

Civitai lists additional base model variants (ZImageTurbo, Qwen, SDXL 1.0) but the Hugging Face release specifically targets FLUX.2 Klein.

How to Use

  1. Load the LoRA via the Load LoRA node
  2. Set strength around 0.8–1.0 (start at 1.0)
  3. Include trigger words in your prompt: Amigurumi, Crochet, a cute bunny plush toy
  4. Generate

Diffusers

from diffusers import FluxPipeline
import torch

pipe = FluxPipeline.from_pretrained(
    "black-forest-labs/FLUX.2-klein-9B",
    torch_dtype=torch.bfloat16
)
pipe.load_lora_weights(
    "artificialguybr/CROCHET-AMIGURUMI-REDMOND-FLUXKLEIN"
)
pipe.to("cuda")

image = pipe(
    "Amigurumi, Crochet, a cute bear plush toy, soft yarn texture",
    num_inference_steps=28,
    guidance_scale=3.5
).images[0]

Amigurumi bear plush with visible stitching detail

Parameter | Value

|---|---

LoRA strength | 0.8–1.0 Steps | 20–30 Guidance (CFG) | 3.0–4.0 Sampler | Euler / DPM++ 2M Resolution | 1024×1024 or 1024×1536 The trigger words Amigurumi, Crochet. should appear at the start of your prompt for best consistency.

Small crochet creature with round features and soft yarn

Common Errors and Fixes

Error | Cause | Fix

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

Style not applying | Trigger words missing or misspelled | Use exactly Amigurumi, Crochet. at prompt start Overcooked look | LoRA strength too high | Reduce to 0.7–0.8 Wrong base model | Using SDXL/other instead of FLUX.2 Klein | Use FLUX.2 Klein 9B base

FAQ

Q: Does this work with SDXL? A: The HF release targets FLUX.2 Klein. Civitai hosts separate variants for SDXL 1.0, ZImageTurbo, and Qwen — download the matching version.

Q: What license applies? A: Apache 2.0 — free for commercial use with attribution.

Q: Recommended LoRA weight? A: Start at 1.0; drop to 0.8 if generations look overstyled.

Sources