Instructions to use bmtdecorai/BMT-FloorPlan-Kontext-LoRA with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use bmtdecorai/BMT-FloorPlan-Kontext-LoRA with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline from diffusers.utils import load_image # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("fill-in-base-model", dtype=torch.bfloat16, device_map="cuda") pipe.load_lora_weights("bmtdecorai/BMT-FloorPlan-Kontext-LoRA") prompt = "Turn this cat into a dog" input_image = load_image("https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/cat.png") image = pipe(image=input_image, prompt=prompt).images[0] - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- Draw Things
🏗️ BMT FloorPlan Kontext LoRA
Convert 2D Architectural Floor Plans to Photorealistic Color Visualizations
📋 Model Overview
|
🎨 Three Visual Styles
|
⚙️ How It Works
This LoRA leverages the FLUX.1 Kontext architecture - a 12B parameter rectified flow transformer trained in latent space with fused DiT blocks and 3D Rotary Positional Embeddings (RoPE). The model has been fine-tuned on a curated dataset of floor plan-to-color render pairs to learn the mapping between 2D CAD layouts and architectural color representations.
🔬 Core Mechanism
graph LR
A[📄 2D Floor Plan<br/>CAD/Blueprint] --> B[🧠 FLUX.1 Kontext<br/>12B Transformer]
B --> C[⚡ LoRA Injection<br/>Low-Rank Weights]
C --> D[🎯 Spatial Understanding<br/>RoPE Encoding]
D --> E[🎨 Style Transfer<br/>B&W / Pure Color / Color Scheme]
E --> F[🏠 Color<br/>Visualization]
📄 Input Conditioning - The model accepts a 2D floor plan image (CAD drawing, architectural sketch, or blueprint) as the primary input.
⚡ LoRA Injection - The fine-tuned low-rank weight matrices are injected into the base FLUX.1 Kontext transformer, specializing the model's attention heads to recognize architectural elements (walls, doors, windows, furniture, spatial layouts).
🎯 Spatial Understanding - Using FLUX.1 Kontext's advanced in-context generation capabilities and RoPE positional encoding, the model interprets the 2D layout as a spatial structure.
🎨 Style Transfer - The LoRA applies learned aesthetic patterns to generate three distinct visual styles: Black & White, Pure Color, and Color Scheme.
🚀 Use Cases
| 🏗️ Architecture | 🏠 Real Estate | 🛋️ Interior Design | 🏗️ Construction | 📚 Education |
|---|---|---|---|---|
| Convert CAD floor plans into client-ready color renders | Generate immersive property previews from technical drawings | Visualize spatial layouts with furniture and material finishes | Create color views for stakeholder presentations | Transform technical drawings into accessible diagrams |
🛠️ Recommended Settings
| Parameter | ⚙️ Recommendation |
|---|---|
| LoRA Strength | 0.9 - 1.0 |
| Base Model | FLUX.1 Kontext (Pro/Max recommended) |
| Resolution | 1024x1024 px (native) or 512x512 px (fast preview) |
| Steps | 30 - 50 |
| CFG Scale | 7.0 - 9.0 |
| Sampler | Euler a / DPM++ 2M Karras |
| Trigger Word | WD |
📝 Prompting Guide
🎯 Basic Structure
WD, floor plan, [style], [room type], [lighting condition]
🎨 Style Keywords
black and white- Monochromatic technical renderpure color- Flat color architectural diagramcolor scheme- Full photorealistic visualization
💡 Example Prompts
# ⚫⚪ Black & White Style
WD, turn this picture into a black and white floor plan.
# 🔵 Deep Blue Map
WD, convert this picture into a deep blue map.
# 🔵 Dark Blue Map
WD, convert this image to a dark blue map.
# 🏠 Standard Floor Plan
WD, turn this picture into a floor plan.
# ☕ Coffee-Colored Style
WD, turn this picture into a coffee-colored plan.
# 🌈 Full Color Scheme
WD, floor plan, color scheme, modern apartment interior,
natural lighting, wooden flooring, white walls, furnished living room and kitchen
⚠️ Limitations
- Resolution Cap: Hard limit at 1024x1024 px; intricate geometry may produce edge artifacts at lower resolutions
- Multi-View Consistency: No built-in guarantee for fully coherent scenes across multiple angles
- Complex Geometry: Highly detailed or irregular architectural elements may require manual refinement
- Style Lock: The model is optimized for the three trained styles; deviations may produce inconsistent results
🔗 Recommended Resources
👤 Model Card Contact
| 👨💻 Creator | Gavin01123 |
| 🌐 Platform | Civitai |
| 📊 Model ID | 1836283 |
| 🔢 Version ID | 2078017 |
| 📅 Joined | July 18, 2025 |
📖 Citation
If you use this model in your research or projects, please cite:
@misc{bmt-floorplan-kontext-lora-2025,
title = {BMT FloorPlan Kontext LoRA},
author = {Gavin01123},
year = {2025},
publisher = {Civitai},
howpublished = {\url{https://huggingface.co/bmtdecorai/BMT-FloorPlan-Kontext-LoRA}}
}
📜 License
This model is subject to the FLUX.1 Kontext license terms by Black Forest Labs. Please refer to the base model license for commercial use restrictions.
The FLUX.1 [dev] Model is licensed by Black Forest Labs, Inc. under the FLUX.1 [dev] Non-Commercial License. Copyright Black Forest Labs, Inc.
⭐ If you find this model helpful, please consider leaving a review on Civitai!
Last updated: August 3, 2025
- Downloads last month
- 18