Best Stable Diffusion Models 2023

by Danesh Ramuthi · Updated March 23, 2023

Stable Diffusion is a text to image model that utilises deep learning. 

Its main purpose is to produce elaborate visual representations based on textual explanations, although it can also be employed for other functions like filling in missing parts of an image, creating new details outside of the original boundaries, and generating translations of images using a text instruction as a guide

As machine learning continues to advance, stable diffusion models have emerged as a promising approach for modelling complex data distributions. 

These models use partial differential equations to describe the flow of information across a high-dimensional space, allowing them to capture long-range dependencies and generate realistic images. 

However, not all diffusion models are created equal, and selecting the best one for a particular task can be a challenging task. 

In this article, we will review the best stable diffusion models for 2023, highlighting their strengths and weaknesses and providing recommendations for different use cases.

Let’s get started

What are Stable Diffusion models?

Stable diffusion models are a type of probabilistic model that is used in machine learning to perform tasks including density estimation, generative modeling, and picture synthesis. These models work by solving a partial differential equation known as the diffusion equation to simulate the dynamics of a high-dimensional system. This equation’s solution represents a stochastic method for generating samples from the target distribution.

Stable diffusion models are well-known for their ability to capture long-term relationships in data distributions and provide high-quality samples. Because of their performance in applications such as image synthesis and generative modeling, they have lately acquired prominence. These models are also stable under specific situations, making them more resistant to fluctuations in the training data and aiding in the prevention of difficulties like mode collapse. Overall, stable diffusion models are a promising strategy in machine learning for modeling complex data distributions.

Where and why can Stable Diffusion Models be used?

Stable Diffusion is a type of latent diffusion model that can produce elaborate visual representations based on textual explanations, and it can also be applied to other functions like filling in missing parts of an image, creating new details outside of the original boundaries, and generating translations of images using a text instruction as a guide.

Its versatility in generating images using different types of prompts makes it a powerful tool with endless possibilities. It has the potential to speed up or enhance the process of creating engaging content for users or customers. Even for creatives who are struggling with a creative block, generated images can serve as inspiration to overcome stagnation. We encourage you to explore the use of Stable Diffusion models to create unique artworks or as a starting point for your artistic or everyday journey.

Best stable diffusion models huggingface

Stability AI - Stable Diffusion Model at Hugging Face

Different stable diffusion versions of Stability AI can be found here at huggingface together with all the relevant files that has been made for public use. 

Stable Diffusion v1-5 Model Card - Huggingface (Runawayml)

Stable Diffusion is a latent model that can generate realistic images based on any given text input. To learn more about the mechanics of Stable Diffusion, refer to the Stable Diffusion blog.

The Stable-Diffusion-v1-5 checkpoint was created by initializing the weights of the Stable-Diffusion-v1-2 checkpoint and then fine-tuning it on “laion-aesthetics v2 5+” for 595k steps at a resolution of 512×512. Additionally, the model was trained with 10% reduced text-conditioning and classifier-free guidance sampling to improve its performance.

The Stable Diffusion model can be used with both the Diffusers library and the RunwayML GitHub repository.

andite/anything-v4.0 - Stable Diffusion Modelc

Fantasy.ai is the sole official hosted AI generation platform that possesses a commercial use license for Anything V4.0. Their service can be accessed at https://Fantasy.ai/.

Any commercial use of Anything V4.0 that is not authorized by Fantasy.ai is strictly prohibited. If you come across any instances of unauthorized commercial use, please report it immediately

CompVis/stable-diffusion-v1-4

Stable Diffusion is a type of latent model that can generate realistic images based on any given text input. To learn more about Stable Diffusion and how it works, check out s Stable Diffusion with Diffusers blog.

The Stable-Diffusion-v1-4 checkpoint was created by initializing the weights of the Stable-Diffusion-v1-2 checkpoint and then fine-tuning it on “laion-aesthetics v2 5+” for 225k steps at a resolution of 512×512. The model was also trained with a 10% reduction in text-conditioning to improve classifier-free guidance sampling.

These weights are intended for use with the Diffusers library. If you’re searching for the weights to be loaded into the CompVis Stable Diffusion codebase, refer to this location.

dreamlike-art/dreamlike-photoreal-2.0

Dreamlike Photoreal 2.0 is a photorealistic model that is based on Stable Diffusion 1.5, and was created by dreamlike.art. 

To make your generated images look more photorealistic, you can add a photo to your prompt. Non-square aspect ratios work better for some prompts. If you want a portrait photo, try using a vertical aspect ratio, and for a landscape photo, use a horizontal aspect ratio.

This model was trained on 768x768px images, so it is recommended to use resolutions such as 768x768px, 640x896px, or 896x640px for optimal performance. However, it also works well with higher resolutions, such as 768x1024px or 1024x768px.

Best Custom (Fine-Tuned) Stable Diffusion Models

Openjourney

A modified version of the Stable Diffusion model called Openjourney attempts to imitate the design of Midjourney. It was made by Prompthero and is freely downloadable and usable by everyone on Hugging Face. With more than 56K downloads in the previous month as of the time of writing, Openjourney is one of the most well-liked fine-tuned Stable Diffusion models on Hugging Face.

Arcane Diffusion

Arcane Diffusion makes an effort to imitate the look of the well-known TV show Arcane. It was made by @nitrosocke and is freely downloadable and usable by everyone on Hugging Face. With more than 30K downloads in the previous month as of this writing, Arcane Diffusion is one of the most well-liked fine-tuned Stable Diffusion models on Hugging Face.

Mo-Di Diffusion

Mo-Di Diffusion makes an effort to imitate the look of contemporary Disney animated movies. It was made by @nitrosocke and is freely downloadable and usable by everyone on Hugging Face. Another well-liked refined Stable Diffusion model is Mo-Di Diffusion, which had more than 26K downloads in the previous month as of the time of writing.

Waifu Diffusion

Model Waifu Diffusion wears anime-style clothing. Everyone can download and use it for free on Hugging Face thanks to the creation of harubaru. With 176K+ downloads in the past month on Hugging Face, Waifu Diffusion is by far the most popular model on this list.

Ghibli Diffusion

Ghibli Diffusion tries to mimic the style of Studio Ghibli. It is created by @nitrosocke and available on Hugging Face for everyone to download and use for free. Ghibli Diffusion is less popular compared to others in this list with 2.8K+ downloads last month on Hugging Face at the time of writing.

Stable Diffusion Model Hash

A stable diffusion model, in the context of hash functions, is a family of hash functions that uses the diffusion process to get a hash value from the input data. The purpose of this method is to diffuse the input data before generating the hash value, making the output not only a function of the input but also a function of the diffusion process. The hash value becomes more unpredictable and challenging to anticipate as a result.

In conclusion, stable diffusion models have emerged as a promising approach for modeling complex data distributions. These models use partial differential equations to describe the flow of information across a high-dimensional space, allowing them to capture long-range dependencies and generate realistic images.

The versatility of stable diffusion models in generating images using different types of prompts makes them a powerful tool with endless possibilities. While there are various stable diffusion models available, the ones reviewed here – Stability AI, anything-v4.0, Stable Diffusion v1-4, and Dreamlike Photoreal 2.0 – have demonstrated impressive results and are suitable for different use cases.

As machine learning continues to advance, it will be exciting to see the future developments and applications of stable diffusion models.

Found that helpful? Get similar posts and other exclusive AI information that we only share with our email subscribers?

Thanks for Reading!

Danesh Ramuthi
CEO

Danesh is a scientist and a content writer with more than 2 years of experience. He is also a published author of a science fiction children’s book titled Imaginary Tales. 

AI has always been in his mind and soul ever since the cult classic movie 2001: A Space Odyssey inspired him to become a writer. Seeing a lot of stigma and misconceptions on AI, he has decided to found Ava Machina as an Hub for people from different backgrounds to gather and learn about AI through expert insights as well as redirecting them to the right source.

Leave a Comment

Your email address will not be published. Required fields are marked *