ArtificialGuyBR

Home / Blog / Training LoRAs

SDXL T-Shirt Design LoRA: TshirtDesignRedmond V2

Generate t-shirt designs with Stable Diffusion XL using the TshirtDesignRedmond V2 LoRA. Trigger word, style tags, settings, and tips.

1 sources cited Training LoRAs

On this page

  1. What the model does
  2. Trigger word and style tags
  3. How to use it
  4. Where it fits in a workflow
  5. Common errors and fixes
  6. FAQ
  7. Sources

SDXL T-Shirt Design LoRA: TshirtDesignRedmond V2

TL;DR:

What the model does

TshirtDesignRedmond V2 is a low-rank adapter (LoRA) built on Stable Diffusion XL 1.0, trained by ArtificialGuyBR on a large dataset of t-shirt design imagery. Instead of generating a generic image of a t-shirt, the LoRA biases the model toward producing actual design artwork the kind of graphic you would print on a shirt: standalone motifs, typography-driven layouts, and clean vector-like compositions.

The author's model card describes it as having "high capacity to generate T-Shirt Designs images," and the training was enabled by GPU time from Redmond.AI. The model is distributed through Hugging Face under the CreativeML OpenRAIL-M license and loads natively with the diffusers library.

Trigger word and style tags

The model responds to the trigger tag TshirtDesignAF. The instance prompt defined on the card is:

TshirtDesignAF, T Shirt Design

From there, the card documents four style keywords that steer the output:

Tag composition is loose: combine the trigger with one style word and the design concept, e.g. TshirtDesignAF, minimalist, mountain logo.

How to use it

Because the repo is a diffusers LoRA adapter, loading it is a standard two-step flow:

from diffusers import DiffusionPipeline

pipe = DiffusionPipeline.from_pretrained("stabilityai/stable-diffusion-xl-base-1.0")
pipe.load_lora_weights("artificialguybr/TshirtDesignRedmond-V2")

prompt = "TshirtDesignAF, T Shirt Design, minimalist wolf head"
image = pipe(prompt).images[0]

The same adapter works in Automatic1111 and ComfyUI: drop the safetensors file into your LoRA folder and add TshirtDesignAF plus your style tag to the prompt.

If you want to try it before installing anything, the author maintains a public demo space, artificialguybr-demo-lora, where all of their LoRAs can be tested in the browser.

Where it fits in a workflow

T-shirt design generation with SDXL typically needs either a strong base-model prompt or a dedicated fine-tune. This LoRA removes most of that prompt engineering: the style lives in the adapter, and the prompt only has to name the subject and the aesthetic. That makes it a practical building block for print-on-demand ideation, mockups, or batch concept exploration.

For related SDXL workflow guides, see related guide and related guide. More LoRAs from the same author are collected at related guide.

Common errors and fixes

Error | Cause | Fix | Source

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

Output looks like stock SDXL, no t-shirt style | Trigger tag TshirtDesignAF missing from prompt | Add TshirtDesignAF, T Shirt Design to the prompt | Model card Adapter has no effect | LoRA weights not loaded, or loaded on a non-SDXL base | Load on stabilityai/stable-diffusion-xl-base-1.0 | Model card base_model tag Results too busy or too plain | No style keyword used | Add detailed, minimalist, colorful, or black and white | Model card

FAQ

What base model does TshirtDesignRedmond V2 need? SDXL 1.0 (stabilityai/stable-diffusion-xl-base-1.0). The adapter is not compatible with SD 1.5 checkpoints.

What is the trigger word? TshirtDesignAF. The full instance prompt is TshirtDesignAF, T Shirt Design.

Can I use the outputs commercially? The model ships under CreativeML OpenRAIL-M, which permits commercial use subject to the license's responsibility terms. Review the full license before shipping products.

Does it work in Automatic1111 and ComfyUI? Yes. It is a standard SDXL LoRA and can be loaded via the LoRA loader in either tool.

Sources