pointblack / stable-interiors-v2

An enhanced version of sd-interior-design

  • Public
  • 66 runs
  • L40S
Iterate in playground

Input

image
*file

Path to input image

*string
Shift + Return to add a new line

Text prompt for design

string
Shift + Return to add a new line

Negative prompt

Default: "lowres, watermark, banner, logo, contactinfo, text, deformed, blurry, blur, out of focus, out of frame, surreal, extra limbs, extra furniture, ugly, overlapping objects, distorted furniture, impossible layout, broken geometry, unrealistic proportions, floating furniture, furniture merging with walls or windows, warped walls, curved walls, mirrors, mirrored surfaces, plush walls, fabric walls, soft focus, poor lighting, artifacts, cut-off furniture"

integer

Number of denoising steps

Default: 50

number

Guidance scale

Default: 15

number

Prompt strength for inpainting

Default: 0.8

integer

Random seed

Output

output
Generated in

Run time and cost

This model runs on Nvidia L40S GPU hardware. We don't yet have enough runs of this model to provide performance information.

Readme

📐 Model Description Enhanced Interior Design Diffusion is a refined and upgraded version of the original sd-interior-design model by Adirik, focused on producing high-quality, diverse interior renderings using Stable Diffusion.

🔧 Key Improvements Base Model Upgrade: Switched from SG161222/Realistic_Vision_V3.0_VAE to a more capable diffusion backbone for enhanced output quality.

VAE Enhancement: Integrated the stabilityai/sd-vae-ft-mse VAE for sharper details, better color accuracy, and reduced artifacts.

High-Resolution Output: Supports 1024×1024 image generation (up from 512×512), delivering much more photorealistic and detailed interiors.

Broader Design Diversity: Enhanced style generalization across various room types and design aesthetics.

🧾 Attribution This model builds upon the original work of sd-interior-design by Adirik. All improvements were made in compliance with the base model’s license.