Stable Diffusion 3 | ALL New Updates Explained & How to Install SD 3 for FREE

AI Catalyst
June 26, 2024
Blog
0 Comments

he latest Stable Diffusion model was released to the public for free, and it’s already creating a lot of buzz. In this article, you will learn how to install Stable Diffusion 3 on your computer and try it yourself. We’ll also explore the new features and improvements in this version and compare it with other AI image-generation tools.

Stable Diffusion 3 Updates

Enhanced Image Quality

SD3 produces high-resolution images up to 2048×2048 pixels, which is significantly higher than previous versions. This improvement is due to its new diffusion architecture, which also reduces image distortion and improves overall image quality.

Better Text Generation

One of the notable upgrades in SD 3 is its ability to generate clearer and more accurate text within images. This was a weakness in earlier versions, but SD3 can now handle text generation on par with leading AI models like DALL-E 3 and Midjourney.

Better Handling of Multi-Subject Prompts

The model has improved its ability to accurately interpret and render images based on prompts involving multiple subjects. This makes it more versatile and capable of creating complex scenes.

Increased Model Sizes

SD 3 offers a range of model sizes from 800 million to 8 billion parameters, meeting the needs of users from hobbyists to professionals. This scalability allows users to choose the right balance between performance and computational requirements.

Speed and Performance

With advanced sampling techniques and powerful hardware, SD3 can generate an image in less than 35 seconds. This speed is crucial for efficient workflow and rapid prototyping.

Safety Measures

Stability AI has prioritized safety in SD3 by implementing different safeguards to prevent the generation of inappropriate or harmful content. This includes the complete removal of NSFW (Not Safe For Work) image generation capabilities.

How to Download and Install Stable Diffusion 3

As of this moment, only Stable Diffusion 3 Medium is available for free download and local use. This is the less powerful model, with only 2 billion parameters. It has already faced a lot of backlash for being quite bad at generating photorealistic images and completely messing up human anatomy. Some users even report that it’s worse than earlier versions like Stable Diffusion 1.5 or SDXL. Nevertheless, if you want to download it and try it yourself, here is a guide on how to do it.

1. Install Python

First, you need to install Python. Open the installer and press Customize Installation. Then click next and check the box for “Add Python to environment variables“. This is very important so don’t forget to do this.

2. Download ComfyUI

Download ComfyUI from Github by pressing the direct download link. Extract the downloaded .7zi file.

3. Download Stable Diffusion 3

Download the 15 GB file from Hugging Face, and the basic workflow file.

4. Open ComfyUI

Open the ComfyUI folder, go to models -> checkpoints and put the 15 GB file inside this folder.

Now, launch run_nvidia_gpu.bat file, which will open ComfyUI in your browser. Drag and drop the workflow file into ComfyUI.

Now drag the wire from Load Checkpoint to Prompt and to Negative prompt and select your model.

Generate Images with SD 3

And this is basically it, you can now generate images with Stable Diffusion 3. Press the Queue Prompt button and your images will be generated.

Stable Diffusion 3 vs Midjourney vs DALL E 3

If you’re wondering how Stable Diffusion 3 performs compared to other AI image generators, here are images generated by SD 3, DALLE 3, and Midjourney using the same prompt.

				
					Stable Diffusion 3

				
					DALL-E 3

				
					Midjourney

				
					Stable Diffusion 3

				
					DALL-E 3

				
					Midjourney

				
					Stable Diffusion 3

				
					DALL-E 3

				
					Midjourney

				
					Stable Diffusion 3

				
					DALL-E 3

				
					Midjourney

				
					Stable Diffusion 3

				
					DALL-E 3

				
					Midjourney

Stable Diffusion 3 - Conclusion

Overall, Stable Diffusion 3 is quite disappointing. The main issue is that the progress isn’t very visible compared to older models, and the developers have only released the weaker, medium version to the public. Photorealistic images are very disappointing, but with proper prompting and experimentation with settings, you can achieve some decent results. The main advantage of Stable Diffusion 3 is that you can use it as a free alternative to Midjourney. Hopefully, it will receive more updates soon and the full model will be released for local use. Whatever happens, we will keep you updated on our YouTube Channel and our website.