Amuse .NET application for stable diffusion

Amuse

github地址：https://github.com/tianleiwu/Amuse

.NET application for stable diffusion, Leveraging OnnxStack, Amuse seamlessly integrates many StableDiffusion capabilities all within the .NET eco-system 

Welcome to Amuse!

Amuse is a professional and intuitive Windows UI for harnessing the capabilities of the ONNX (Open Neural Network Exchange) platform, allowing you to easily augment and enhance your creativity with the power of AI.

Amuse, written entirely in .NET, operates locally with a dependency-free architecture, providing a secure and private environment and eliminating the need for intricate setups or external dependencies such as Python. Unlike solutions reliant on external APIs, Amuse functions independently, ensuring privacy by operating offline. External connections are limited to the essential process of downloading models, preserving the security of your data and shielding your creative endeavors from external influences.

Experience the power of AI without compromise

Features

• Paint To Image: Experience real-time AI-generated drawing-based art with stable diffusion.
• Text To Image: Generate stunning images from text descriptions with AI-powered creativity.
• Image To Image: Transform images seamlessly using advanced machine learning models.
• Image Inpaint: Effortlessly fill in missing or damaged parts of images with intelligent inpainting.
• Model Management: Install, download and manage all your models in a simple user interafce.

Amuse provides compatibility with a diverse set of models, including

• StableDiffusion 1.5
• StableDiffusion Inpaint
• SDXL
• SDXL Inpaint
• SDXL-Turbo
• LatentConsistency
• LatentConsistency XL
• Instaflow

Why Choose Amuse?

Amuse isn't just a tool; it's a gateway to a new realm of AI-enhanced creativity. Unlike traditional machine learning frameworks, Amuse is tailored for artistic expression and visual transformation. This Windows UI brings the power of AI to your fingertips, offering a unique experience in crafting AI-generated art.

Key Highlights

• Intuitive AI-Enhanced Editing: Seamlessly edit and enhance images using advanced machine learning models.
• Creative Freedom: Unleash your imagination with Text To Image, Image To Image, Image Inpaint, and Live Paint Stable Diffusion features, allowing you to explore novel ways of artistic expression.
• Real-Time Results: Witness the magic unfold in real-time as Amuse applies live inference, providing instant feedback and empowering you to make creative decisions on the fly.

Amuse is not about building or deploying; it's about bringing AI directly into your creative process. Elevate your artistic endeavors with Amuse, the AI-augmented companion for visual storytellers and digital artists.

Paint To Image

Paint To Image is a cutting-edge image processing technique designed to revolutionize the creative process. This method allows users to paint on a canvas, transforming their artistic expressions into high-quality images while preserving the unique style and details of the original artwork. Harnessing the power of stable diffusion, Paint To Image opens up a realm of possibilities for artistic endeavors, enabling users to seamlessly translate their creative brushstrokes into visually stunning images. Whether it's digital art creation, stylized rendering, or other image manipulation tasks, Paint To Image delivers a versatile and intuitive solution for transforming painted canvases into captivating digital masterpieces.

Text To Image

Text To Image Stable Diffusion is a powerful machine learning technique that allows you to generate high-quality images from textual descriptions. It combines the capabilities of text understanding and image synthesis to convert natural language descriptions into visually coherent and meaningful images

Image To Image

Image To Image Stable Diffusion is an advanced image processing and generation method that excels in transforming one image into another while preserving the visual quality and structure of the original content. Using stable diffusion, this technique can perform a wide range of image-to-image tasks, such as style transfer, super-resolution, colorization, and more

Image Inpaint

Image inpainting is an image modification/restoration technique that intelligently fills in missing or damaged portions of an image while maintaining visual consistency. It's used for tasks like photo restoration and object removal, creating seamless and convincing results.

Model Manager

Discover the simplicity of our Model Manager – your all-in-one tool for stress-free model management. Easily navigate through an intuitive interface that takes the hassle out of deploying, updating, and monitoring your stable diffusion models. No need for configuration headaches; our Model Manager makes it a breeze to install new models. Stay in control effortlessly, and let your creative process evolve smoothly.

Getting Started
Get started now with our helpful documentation: https://github.com/Stackyard-AI/Amuse/blob/master/Docs/GettingStarted.md

Hardware Requirements

Compute Requirements

Generating results demands significant computational time. Below are the minimum requirements for accomplishing such tasks using Amuse

Device	Requirement
CPU	Any modern Intel/AMD
AMD GPU	Radeon HD 7000 series and above
Intel	HD Integrated Graphics and above (4th-gen core)
NVIDIA	GTX 600 series and above.

Memory Requirements

AI operations can be memory-intensive. Below is a small table outlining the minimum RAM or VRAM requirements for Amuse

Model	Device	Precision	RAM/VRAM
Stable Diffusion	GPU	16	~4GB
Stable Diffusion	CPU/GPU	32	~8GB
SDXL	CPU/GPU	32	~18GB

System Requirements

Amuse provides various builds tailored for specific hardware. DirectML is the default choice, offering the broadest compatibility across devices.

Build	Device	Requirements
CPU	CPU	None
DirectML	CPU, AMD GPU, Nvidia GPU	At least Windows10
CUDA	Nvidia GPU	CUDA 11 and cuDNN toolkit
TensorRT	Nvidia GPU	CUDA 11 , cuDNN and TensorRT libraries

Realtime Requirements

Real-time stable diffusion introduces a novel concept and demands a substantial amount of resources. The table below showcases achievable speeds on commonly tested graphics cards

Device	Model	FPS
GTX 2080	LCM_Dreamshaper_v7_Olive_Onnx	1-2
RTX 3090	LCM_Dreamshaper_v7_Olive_Onnx	3-4