1

14t6

Developed by ismot
A diffusion-based text-to-image generation model that supports image generation and modification through text prompts
Downloads 24
Release Time : 1/14/2023

Model Overview

Stable Diffusion v2-1 is a further fine-tuned latent diffusion model based on the v2 version, using the OpenCLIP-ViT/H text encoder and supporting high-resolution image generation

Model Features

High-resolution generation
Supports 768x768 resolution image generation with improved quality over previous models
Safe training
Uses a conservative threshold of p_unsafe=0.1 to filter training data, reducing harmful content generation
Multi-framework support
Compatible with the diffusers library and native stablediffusion codebase
Fine-tuning optimization
Additional training of 210k steps (55k conservative steps + 155k open steps) beyond the v2 version

Model Capabilities

Text-to-image generation
Image modification
Art creation
Design assistance

Use Cases

Art creation
Concept art generation
Quickly generates concept art based on text descriptions
Can produce concept designs in various styles
Education and research
Generative model research
Used to study the limitations and biases of generative models
Creative tools
Design prototype generation
Quickly generates product design prototypes
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
© 2025AIbase