Introduction
You might hear of Stable diffusion a lot recently. Like Midjourney, it is an innovative technology that has recently gained attention for its impressive potential in art generation. In this blog post, we’ll introduce the basics of stable diffusion and how can we access its service.
What is Stable Diffusion?
Stable diffusion is an open-source Python program released in August 2022 to generate art. It was developed together by researchers and start-up companies, notably Stability AI, which has raised millions of dollars in funds.
It took a vast amount of computing resources (256 NVidia GPU) and 600,000 US dollars to train the initially released model according to Stability AI’s founder Emad Mostaque. Along with the Python program, the trained model is also open-sourced, which really makes it thrives.
Training a model is not cheap, but deploying a trained model is not very hard. A personal PC with an 8G GPU shall be able to serve as a Stable Diffusion art generator. This makes a lot of companies or individuals can provide image-generation services to others.
We’ll discuss how to deploy our own Stable diffusion service in another post. In this article, we’ll look through a few nice service providers.
Services Provided by Stability AI
Let’s first take a look at a few services provided by one of its creators, Stability AI.
The official Demo on Hugging Face
The latest 2.1 demo is hosted on Huggine Face. You may input a prompt and a negative prompt in the second line.
The original version 1 demo is also available online. But it’s not popular as version 2 of course.
Dream Studio
Dream Studio is the official image generator by Stability AI. It provides access to more settings, such as the ratio aspect, prompt strength, etc. As of Apr 2023, you can choose between Stable Diffusion 2.1 and the latest SDXL Beta. Dream Studio is not free after you use up the free credits.
Third-party Online Service
Playground AI
If you are familiar with the way Midjourney works, I recommend using playground AI. It’s backed up by a dedicated team, and you can create up to 1,000 images per day without any delay. You may also switch between Stable Diffusion v1.5 and v2.1. Alternatively, you can purchase DALL-E 2 add-ons powered by Open AI. Playground also offers its own v1 model.
stablediffusionweb.com
The image generator provided by stablediffusionweb.com is not that attractive, but it promises that they do not store your text prompts and images.
The prompts database and generator from this website could be very handy when you need some inspiration.
Crowdsourced Cluster
Unlike the above Online Service with the computing power, especially GPU consumption, provided by companies behind the service, AI horde provides a way that these computing resources are provided by crowdsource distributed clusters provided by like you and me.
Please check out Stable UI and artificial art for some fun experiments. Along with Stable Diffusion, they offer more than a handful of other image-generation models.
Web Service Built upon Stable Diffusion
There are quite a few projects, notably Stable Diffusion web UI and Stable Diffusion UI v2, offering a one-click setup for a full web UI.
There are demos of Stable Diffusion Web UI on Hugging Face, which can be accessed here.
Photoshop Plug-in
This is another great example of how open source promotes collaboration, ingenuity, and swift progress. Stability AI offers the Stability plugin for Photoshop. There are also open-source plug-ins available such as stable.art and Auto-Photoshop-StableDiffusion-Plugin.
Conclusion
The community surrounding the Stable Diffusion is really thriving and technologies still evolve quickly every day. If you are a programmer, you’ll need to catch up with the latest trends; if you are an artist, start to use AI art to facilitate your design. I still firmly believe that AI-generated art is a tool for real artists.
If you are interested in more AI art, you are welcome to visit our Midjourney tutorial. Please feel free to leave your comments below.