ArtificialGuyBR

Home / Blog / Training LoRAs

Doodle Style LoRA: Hand-Drawn Images with Z-Image Turbo

How to use the Doodle Redmond LoRA for Z-Image Turbo: trigger words, ComfyUI setup, and the hand-drawn doodle aesthetic.

1 sources cited Training LoRAs

On this page

  1. What the Doodle Redmond LoRA does
  2. Trigger words and prompt setup
  3. How to use it in ComfyUI
  4. License and usage rights
  5. Common errors and fixes
  6. FAQ
  7. Sources

Doodle Style Redmond LoRA for Z-Image Turbo

TL;DR:

What the Doodle Redmond LoRA does

The Doodle Redmond ZIMAGE LoRA from ArtificialGuyBR is a small adapter trained on doodle-style images. It shifts the base model's output toward a hand-drawn, sketch-like look: simple shapes, clear outlines, and playful subjects, with consistent styling across a generation. According to the model card, it produces "high-quality doodle style content with excellent consistency" and keeps a high level of detail.

The name comes from the training sponsorship: the model was trained with GPU time from Redmond.AI. The LoRA itself is trained on top of Tongyi-MAI/Z-Image-Turbo, a fast text-to-image base model, which is why the release is tagged as a Z-Image-Turbo adapter.

Doodle style sample generated with the Doodle Redmond LoRA

Example gallery output included in the model card, 1024x1024 generation using the Doodle / DoodleRedm trigger tokens.

Trigger words and prompt setup

The model card lists the instance prompt as Doodle, DoodleRedm and instructs you to use both tokens to activate the style. A typical prompt looks like:

Doodle, DoodleRedm, a cute robot mascot, thick outline, flat colors

Keep the trigger words near the start of the positive prompt. As with most style LoRAs, adding descriptive terms for your subject after the trigger tokens gives you the doodle treatment applied to whatever you describe.

Second doodle style sample from the model card

Another gallery image from the model card showing the consistent hand-drawn look across different subjects.

How to use it in ComfyUI

The author recommends ComfyUI for best results. Setup is a standard LoRA workflow:

  1. Download the Safetensors weight from the Files tab and place it in your ComfyUI models/loras folder.
  2. Add a Load LoRA node and connect it between your checkpoint model and the sampler.
  3. Set the LoRA name to the downloaded file and start with a strength around 1.0.
  4. Put Doodle, DoodleRedm in the positive prompt and generate at 1024x1024, the resolution of the sample gallery.

If you are new to LoRA workflows, see the general LoRA setup guide for diffusers and ComfyUI basics, or browse other Z-Image Turbo LoRAs in the portfolio.

The LoRA also works with the diffusers pipeline. Load Z-Image Turbo with DiffusionPipeline and use the load_lora_weights method with the downloaded weight file, then prompt with the same trigger tokens.

Third doodle style sample showing style range

Third gallery sample from the model card, showing the style applied to a different subject while keeping the same outline-and-flat-color aesthetic.

License and usage rights

The model is released under the Apache License 2.0, so you can use, modify, and redistribute it, including in commercial projects, as long as you follow the license terms. Unlike several other style LoRAs in the portfolio, there is no non-commercial restriction here.

The author asks for support via Patreon, Ko-fi, or Buy Me a Coffee if you find the model useful, but that is optional and not required for use.

Common errors and fixes

Error | Cause | Fix | Source

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

Style not appearing | Trigger words missing from prompt | Add Doodle, DoodleRedm to the positive prompt | Model card trigger words No effect after loading | LoRA node connected after sampler | Place the Load LoRA node before the sampler in the chain | Author's ComfyUI recommendation Wrong base model mismatch | Loaded on a non-Z-Image Turbo checkpoint | Use Tongyi-MAI/Z-Image-Turbo as the base model | Model card base_model field

FAQ

What base model does this LoRA need?

Tongyi-MAI/Z-Image-Turbo, listed as the base model in the model card. It is a diffusers text-to-image model.

What are the trigger words?

Doodle and DoodleRedm. Both are listed in the model card, and the instance prompt is Doodle, DoodleRedm.

Is the model free to use commercially?

The weights are Apache 2.0 licensed, so yes, under the terms of that license.

What resolution should I generate at?

The sample gallery is 1024x1024, matching the native training resolution of the base model.

Why is it called Redmond?

Redmond.AI sponsored the GPU time used to train the model, per the model card.

Sources