S

Stable Diffusion V1 5 Inpainting

Developed by botp
A text-to-image generation model based on latent diffusion architecture with enhanced image inpainting capabilities using masks
Downloads 6,191
Release Time : 5/5/2023

Model Overview

This model can not only generate realistic images from text inputs but also intelligently repair images through masks. Initialized with Stable-Diffusion-v-1-2 weights, it adds 5 input channels specifically for processing mask information.

Model Features

Dual Functionality
Supports both text-to-image generation and mask-based image inpainting
Enhanced Training
Additional 440K steps of inpainting-specific training on LAION dataset, optimized with 10% text condition dropout
Mask Processing Optimization
UNet incorporates 5 dedicated input channels, with 25% training samples using full masks for enhanced robustness

Model Capabilities

Text-guided image generation
Image inpainting and editing
High-resolution image synthesis
Artistic creation assistance

Use Cases

Creative Design
Concept Art Generation
Rapidly generate design concept images from text descriptions
512x512 resolution images with support for iterative refinement
Image Editing
Intelligent Photo Retouching
Automatically repair photo defects or remove unwanted elements through masks
FID 1.00, LPIPS 0.141 (outperforms specialized inpainting models like LaMa)
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
© 2025AIbase