ArtificialGuyBR

Home / Blog / Training LoRAs

Coloring Book LoRA for Qwen Image 2512 by ArtificialGuyBR

A Qwen/Qwen-Image-2512 LoRA that generates clean line art coloring book illustrations. Learn trigger words, usage, and recommended settings.

2 sources cited Training LoRAs

Coloring Book LoRA for Qwen Image 2512 by ArtificialGuyBR

TL;DR

What It Does

ColoringBook Redmond QwenImage is a LoRA (Low-Rank Adaptation) model fine-tuned for the Qwen/Qwen-Image-2512 diffusion base. It steers generation toward coloring book-style illustrations: clean, single-weight black lines on a white background, minimal interior detail, and simple subjects — the kind of art ready to be colored in by hand or digitally.

Coloring book line art sample

Training and Base Model

The model was trained on a curated dataset of high-quality coloring book images. The author credits Redmond.AI for providing the GPU time used during fine-tuning. Per the model card, the training objective emphasizes consistency so that output keeps a uniform line weight and stylistic coherence across varied prompts.

Trigger Words

To activate the style, include either Coloring Book or ColoringBookAF in your generation prompt. Both tokens are recognized trigger words. Prefixing the prompt with one of these terms is the most reliable way to engage the LoRA.

The author recommends ComfyUI for best results. Load the LoRA as a standard LoRA node in a Qwen-Image-2512 workflow. While specific CFG scale, step count, and resolution values are not documented by the author, a good starting point mirrors standard Qwen Image practice: moderate guidance (5–10), 20–30 steps, and a 1024×1024 canvas.

To reinforce the line-art aesthetic in your prompt, consider appending descriptors such as:

Examples

Coloring book illustration sample

The samples show the model's output: clean line illustrations with consistent stroke weights and subjects suitable for coloring.

License

ColoringBook Redmond QwenImage is released under the Apache License 2.0, allowing both commercial and non-commercial reuse with attribution.

Sources