ArtificialGuyBR

Home / Blog / Training LoRAs

3DRedmond V1: SDXL LoRA for 3D Render Style Images

LoRA for SDXL that generates stylized 3D render images with clean lighting. Trigger words, usage guide, and example outputs.

1 sources cited Training LoRAs

On this page

  1. What This Model Does
  2. Trigger Words & Prompting
  3. Recommended Settings
  4. Example Outputs
  5. How to Use
  6. License
  7. Common Errors and Fixes
  8. FAQ
  9. Sources
  10. Related Models

3DRedmond V1: SDXL LoRA for 3D Render Style Images

TL;DR

What This Model Does

3DRedmond V1 is a Low-Rank Adaptation (LoRA) trained on SDXL 1.0 that steers generations toward a stylized 3D render look — think clean lighting, smooth surfaces, and a distinct "rendered" quality rather than photorealism. The author describes it as "the ultimate LORA for creating stunning 3D Render Style Images."

The model was trained on a large dataset with GPU time sponsored by Redmond.AI. It's distributed as a Safetensors file (~200-500 MB typical for SDXL LoRAs) and loads into any SDXL pipeline.

Trigger Words & Prompting

You must include both trigger tokens in your prompt:

3D Render Style, 3DRenderAF, <your subject description>

The model card's example prompts:

Recommended negative prompt:

bad art, ugly, deformed, watermark, text

Parameter | Value

|---|---

Base model | stabilityai/stable-diffusion-xl-base-1.0 LoRA weight | 0.7–1.0 (start at 0.8) Sampler | DPM++ 2M Karras / Euler a Steps | 20–30 CFG | 5–7 Resolution | 1024×1024 (SDXL native) The example images on the model card were generated at 1024×1024.

Example Outputs

Cute panda on a table in a kitchen, 3D render style with clean lighting and smooth surfaces

Portrait of a beautiful woman in stylized 3D render style, smooth skin and clean lighting

Portrait of a humanoid fox, anthropomorphic character with fur detail in 3D render aesthetic

All three samples show the model's consistent style: stylized 3D renders with smooth shading, clean specular highlights, and a cohesive "rendered" look that's distinct from both photorealism and 2D illustration.

How to Use

Diffusers (Python)

from diffusers import StableDiffusionXLPipeline
import torch

pipe = StableDiffusionXLPipeline.from_pretrained(
    "stabilityai/stable-diffusion-xl-base-1.0",
    torch_dtype=torch.float16,
    variant="fp16"
).to("cuda")

pipe.load_lora_weights("artificialguybr/3DRedmond-V1")

prompt = "3D Render Style, 3DRenderAF, Portrait of a cute robot, cinematic lighting"
image = pipe(prompt, num_inference_steps=25, guidance_scale=6).images[0]

Automatic1111 / ComfyUI

  1. Download the .safetensors file from the Files tab
  2. Place in models/Lora/ (A1111) or models/loras/ (ComfyUI)
  3. Add &lt;lora:3DRedmond-V1:0.8&gt; to your prompt (or use the LoRA loader node in ComfyUI)
  4. Include both trigger words: 3D Render Style, 3DRenderAF

License

The model card does not specify an explicit license. The base model (SDXL 1.0) is under the Stability AI Community License. Check the model page for updates before commercial use.

Common Errors and Fixes

Error | Cause | Fix

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

Style not applying | Missing trigger words | Always include both 3D Render Style and 3DRenderAF Overcooked / burned look | LoRA weight too high | Reduce weight to 0.6–0.7 Blurry / low detail | Too few steps | Use 25–30 steps with DPM++ 2M Karras Wrong base model | Using SD 1.5 or Pony | Must use SDXL 1.0 base (stabilityai/stable-diffusion-xl-base-1.0)

FAQ

Q: Does this work with SD 1.5? A: No. This is an SDXL LoRA and requires an SDXL base model.

Q: Can I use it with Pony Diffusion XL? A: It may load but results will be unpredictable. The LoRA was trained on the official SDXL 1.0 base.

Q: What LoRA weight should I use? A: Start at 0.8. Go lower (0.6) if the style overwhelms your prompt; go higher (1.0) for maximum style transfer.

Q: Are the trigger words case-sensitive? A: The model card shows them capitalized. Match the casing exactly for reliable activation.

Sources