ArtificialGuyBR

Home / Blog / Fixing things

Digital Asset REDMOND LoRA for FLUX.2 Klein 9B

LoRA for FLUX.2 Klein 9B that generates Digital Asset Media style images with high consistency. Trigger word: Digital Asset.

2 sources cited Fixing things

On this page

  1. What this model does
  2. Trigger word
  3. How to use
  4. Recommended settings
  5. Model variants (Civitai)
  6. License
  7. Common errors and fixes
  8. FAQ
  9. Sources

Digital Asset REDMOND LoRA for FLUX.2 Klein 9B

TL;DR

What this model does

Digital Asset REDMOND is a Low-Rank Adaptation (LoRA) trained on top of black-forest-labs/FLUX.2-klein-9B. It specializes in generating Digital Asset Media style images — clean, high-detail digital illustrations suited for game assets, UI elements, iconography, and similar media. The training compute was sponsored by Redmond.AI.

The model produces consistent outputs when the trigger phrase is used, making it reliable for batch generation of stylistically coherent assets.

Digital Asset REDMOND example 1

Trigger word

Prepend your prompt with:

Digital Asset.

The period is part of the trigger. Example:

Digital Asset. isometric game asset, treasure chest, gold trim, clean background

How to use

  1. Download the .safetensors file from the Hugging Face repository.
  2. Place it in ComfyUI/models/loras/.
  3. Add a Load LoRA node, select the file, and set strength (model weight) to 0.7–1.0.
  4. Connect it to your FLUX.2 Klein 9B checkpoint.
  5. Include Digital Asset. at the start of your prompt.

diffusers (Python)

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/DIGITALASSET-REDMOND-FLUXKLEIN")
pipe.to("cuda")

image = pipe(
    "Digital Asset. isometric sword, fantasy RPG style, clean edges",
    num_inference_steps=28,
    guidance_scale=3.5
).images[0]

Automatic1111 / WebUI Forge

Setting | Value

|---|---

Base model | FLUX.2 Klein 9B (black-forest-labs/FLUX.2-klein-9B) LoRA weight | 0.7 – 1.0 Sampler | Euler / Euler a Steps | 20 – 30 Guidance (CFG) | 3.0 – 4.0 Resolution | 1024×1024 (native) Higher LoRA weights (0.9–1.0) increase style adhesion but may reduce prompt adherence. Start at 0.8.

Digital Asset REDMOND example 2

Model variants (Civitai)

On Civitai the same LoRA is packaged for three base-model ecosystems:

All share the same trigger word (Digital Asset.) and training data.

License

Apache License 2.0 — free for commercial use, modification, and distribution with attribution. See the HF model card for the full license text.

Common errors and fixes

Error | Cause | Fix

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

Style not appearing | Missing or misspelled trigger word | Ensure prompt starts with Digital Asset. exactly (with period) Over-saturated / burnt look | LoRA weight too high | Reduce weight to 0.6–0.7 Prompt ignored | LoRA weight too high | Lower weight; increase guidance scale slightly Wrong base model | Using non-FLUX checkpoint | Use FLUX.2 Klein 9B or matching variant (ZImageTurbo / Qwen Image)

FAQ

Q: Can I use this with FLUX.1-dev or FLUX.1-schnell? A: Not recommended. The LoRA was trained specifically on FLUX.2 Klein 9B. Cross-base-model usage typically degrades quality.

Q: What resolution works best? A: 1024×1024 is the native training resolution. FLUX supports arbitrary aspect ratios; 1024×1024 or 768×1344 (portrait) work well.

Q: Is the trigger word case-sensitive? A: The training used Digital Asset. with capital D and A. Match that casing for best results.

Q: Where can I download the model? A: Hugging Face (Files tab) or Civitai.

Sources