Browse Models

SG_161222 /

Realistic Vision

Family

Stable Diffusion 1

Type

Fine-Tuned Model

License

CreativeML Open RAIL-M License

Released

2024-05-12

How To Use

Lab Station OS

The simplest way to self-host Realistic Vision. Launch a dedicated cloud GPU server running Lab Station OS to download and serve the model using any compatible app or framework.

Direct Download

Download model weights for local inference. Must be used with a compatible app, notebook, or codebase. May run slowly, or not work at all, depending on your system resources, particularly GPU(s) and available VRAM.

Browse Compatible Apps

comfyanonymous /

ComfyUI

Generate images and videos using a powerful low-level workflow graph builder - the fastest, most flexible, and most advanced visual generation UI.

lllyasviel /

Stable Diffusion WebUI Forge

Forge is a platform built on top of Stable Diffusion WebUI to make development easier, optimize resource management, speed up inference, and study experimental features.

Automatic1111 /

Stable Diffusion Web UI

Automatic1111's legendary web UI for Stable Diffusion, the most comprehensive and full-featured AI image generation application in existence.

bmaltais /

Kohya's GUI

Train your own LoRAs and finetunes for Stable Diffusion and Flux using this popular GUI for the Kohya trainers.

Model Report

SG_161222 / Realistic Vision

Realistic Vision V6.0 B1 is a Stable Diffusion checkpoint model trained on 3,000 images over 664,000 steps. It excels at portrait and full-body image generation, using a merged architecture from multiple models. Best results achieved with DPM samplers at CFG 1.5-2.0 and lower resolutions.

Explore the Future of AI

Your server, your data, under your control

Model Architecture and Development

The latest version, V6.0, comes in two variants: B1 and B2. The B1 variant was trained on approximately 3,000 images over 664,000 steps, while B2 represents a full re-training with an expanded dataset of 6,400 images and 724,000 training steps. The model incorporates elements from numerous other models, including HassanBlend, URPM, and Protogen, among others.

The model is available both with and without a Variational Autoencoder (VAE), though for V6.0, using the stabilityai/sd-vae-ft-mse-original VAE is recommended for optimal results.

Features and Capabilities

Realistic Vision excels at generating photorealistic images across various scenarios, with particular strength in portrait generation. The model supports multiple resolution configurations:

Face portraits: 512×512 or 640×640

Portraits: 512×768

Half body: 512×768

Full body: 512×768 or 640×960

Version 6.0 brought significant improvements in generation resolution and enhanced capabilities for both NSFW and SFW content, particularly in female anatomy rendering. However, some challenges remain with pose generation at higher resolutions, which are being addressed in ongoing development.

Optimal Usage and Settings

For best results, the creator recommends:

Samplers: DPM++ SDE Karras or DPM++ 2M SDE

CFG Scale: 1.5-2.0

Hires.Fix for full-body and half-body images

ADetailer model for enhanced results

The model benefits from specific negative prompts to avoid common issues like anatomical errors and unrealistic elements. For high-resolution outputs, the recommended workflow includes:

Denoising steps: 20-30

Denoising strength: 0.4

Upscaler: 4x-UltraSharp

The model is licensed under CreativeML Open RAIL++-M, allowing for both personal and commercial use within specified guidelines.

References

Civitai Model Page - Main model repository and documentation

Hugging Face Repository - Alternative download source and technical documentation

CreativeML Open RAIL++-M License - Full license text

Tutorial Video - Detailed usage guide

Detail Tweaker LoRA - Recommended complementary model

4x-UltraSharp Upscaler - Recommended upscaling solution