ArtificialGuyBR

Home / Blog / Training LoRAs

Crochet Amigurumi LoRA for Qwen Q-Image 2512

LoRA adapter for Qwen Q-Image 2512 that adds a crochet amigurumi style to generated images, with trigger words and recommended settings.

2 sources cited Training LoRAs

On this page

  1. What the model does
  2. Credits and training
  3. How to use it
  4. License
  5. Common errors and fixes
  6. FAQ
  7. Sources

Crochet Amigurumi LoRA for Qwen Q-Image 2512

Crochet amigurumi character generated with the Redmond LoRA

What the model does

The Crochet Amigurumi Redmond LoRA is a low-rank adaptation (LoRA) trained on the Qwen Q-Image 2512 base model. It steers generations toward the crochet amigurumi aesthetic: small stitched characters, soft fiber textures, and the distinctive round-headed plush look of handmade amigurumi.

Credits and training

The author, ArtificialGuyBR, acknowledges GPU time donated by Redmond.AI for training. The instance prompt baked into the model card is Amigurumi, Crochet., and the published trigger words are Amigurumi and Crochet..

How to use it

  1. Download the Safetensors file (≈ 81 MB) from the Hugging Face repo.
  2. Load it into your Qwen Q-Image 2512 pipeline. The model card recommends ComfyUI for the best results — load the LoRA node and connect it to the Qwen base model.
  3. Include Amigurumi and Crochet. in your prompt to activate the style.
  4. Tune the LoRA strength. Around 0.6–0.8 blends the style gently; 0.8–1.0 pushes the crochet texture more aggressively.

License

Apache 2.0 — commercial and derivative use are permitted. Internal link for related craft LoRAs.

Common errors and fixes

Error | Cause | Fix | Source

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

LoRA has no effect | Trigger words missing from prompt | Add Amigurumi and Crochet. | HF card Model not loading | Wrong base model selected | Must use Qwen Q-Image 2512, not Flux/KLEIN variants | Civitai Weak texture | LoRA strength too low | Raise to 0.8–1.0 | Author rec

FAQ

Q: Can I use this with AUTOMATIC1111? A: The card recommends ComfyUI, but any Qwen Q-Image 2512 diffusers pipeline that supports Safetensors LoRAs should work.

Q: Is the base model Qwen Q-Image 2512? A: Yes — the HF card lists base_model: Qwen/Qwen-Image-2512. Civitai also offers Zimage, Flux, and KLEIN variants; pick the Qwen one.

Sources