How to Use Stable Diffusion

Codecademy Team
Learn how to generate images using Stable Diffusion.


The evolution of Artificial Intelligence (AI) has been remarkable, progressing from early stages with chatbots like ELIZA to its current application in generating images. Today, AI serves as an invaluable tool in our daily lives. One of the latest advancements in generative AI is Stable Diffusion, a cutting-edge technology capable of generating images based on both textual and imagery-based prompts.

Stable Diffusion has emerged as a leading contender in the AI image generation landscape. Its proficiency lies in producing realistic faces from concise prompts while offering extensive customization options during the image generation process. This article aims to delve into the capabilities of Stable Diffusion, providing a comparative analysis against other AI image generation tools.

Stable Diffusion Image Generation

Stable Diffusion is a unique and complex AI image generator. It can modify each aspect of an image, or create a completely new image from a source image. However, users may find the interface difficult to use at first. Before we dive deep into understanding Stable Diffusion as a tool, let’s take a look at how to access Stable Diffusion.

Stable Diffusion stands out as a sophisticated and versatile AI image generator, capable of manipulating various aspects within an image, augmenting existing content, or creating new visual compositions from source images. For new users, its interface may initially present some difficulty. Before exploring Stable Diffusion’s functionalities, let’s explore how to access and utilize this tool effectively.

Accessing Stable Diffusion

Developed by Stability AI, Stable Diffusion is accessible through the suite of AI tools known as Clipdrop available at Clipdrop. However, note that access to Stable Diffusion via Clipdrop requires the Pro version, which is a paid subscription.

Stable Diffusion via Clipdrop is no longer free. It requires the Pro version of Clipdrop.

Image of Clipdrop Pricing Tiers

Stable Diffusion is an open-source product. Other websites, such as Stable Diffusion Web, integrate Stable Diffusion to provide features to their users. Stable Diffusion Web offers image generation features, albeit with fewer options compared to Clipdrop.

Pros and Cons of Stable Diffusion

One of the most remarkable aspects of Stable Diffusion is its utilization of Negative Prompts. Negative Prompts enable users to specify what elements to avoid during image generation, preventing unintended image features.

Moreover, Stable Diffusion boasts an extensive array of customization options. Users can modify original images or generate entirely new images based on a source image. Leveraging the comprehensive tools within Clipdrop, users can crop, refine, adjust lighting, and access numerous other features embedded in Stable Diffusion.

However, there are some drawbacks to consider. Utilizing Stable Diffusion via the Stability AI website incurs a minimum cost of $13(USD) per month. Additionally, navigating the interface can be somewhat challenging, requiring users to transition between different tools to apply additional effects.

How to Use Stable Diffusion

We will walk through an example of how to use Stable Diffusion. Let’s use ChatGPT to create an image prompt for Stable Diffusion to create an image of a dog and a cat lying together near a fireplace that we can later download.

Here is the prompt generated by ChatGPT:

A heartwarming and cozy scene featuring a dog and a kitten peacefully lying together on a soft mat in front of a crackling fireplace inside a charming house. The room should exude warmth and comfort, with soft lighting and inviting decor. The dog and kitten should appear content and relaxed, showcasing a genuine bond. Capture the essence of tranquility and companionship in this domestic setting. Pay attention to details like the texture of the mat, the flickering flames in the fireplace, and the overall cozy atmosphere 

To place this prompt in Stable Diffusion, access the platform through Clipdrop (Pro version required).

Image of Stable Diffusion Home Page Prompts

Upon inputting the prompt, Stable Diffusion generates four images, each portraying the described scene.

Stable Diffusion Generated Image 1 Stable Diffusion Generated Image 2 Stable Diffusion Generated Image 3 Stable Diffusion Generated Image 4

Most of the images look really good! There is one image that looks a bit strange, do you see it?


The first image generated a dog with a few extra limbs! Remember, generative AI isn’t capable of “thinking” like a human. It doesn’t inherently know that a dog only has four limbs. Hopefully, it was trained with images of dogs that only have four limbs.

We’ll use Stable Diffusion’s unique feature to modify these images. First, we press the back button at the top left. Then, we’ll add a negative prompt to minimize some of the deficiencies, such as the extra limbs in the first image.

Image of Using Stable Diffusion Negative Prompt

Here are the resulting images:

Stable Diffusion Generated Image with Negative Prompt 1 Stable Diffusion Generated Image with Negative Prompt 2 Stable Diffusion Generated Image with Negative Prompt 3 Stable Diffusion Generated Image with Negative Prompt 4

Not perfect, but definitely better! I know it seems strange but saying “extra limbs; worst quality; ugly;” are indicating the features to avoid during image generation.

Stable Diffusion is not only capable of image generation, but it can also edit our images. Once an image is generated, we can click the Edit button at the top right to access any of the following features:

  • Remove Background
  • Cleanup Imperfections
  • Relight
  • Upscale
  • Reimagine
  • Uncrop
  • Replace Background
  • Sky Replacer

Let’s take a look at how Stable Diffusion compares to the other AI image generators.

Other Image Generators

While Stable Diffusion showcases remarkable capabilities in image generation, there are other powerful AI generators like DALL-E and Midjourney. Let’s examine the distinguishing features and functionalities that set these three tools apart.

DALL-E Midjourney Stable Diffusion
Training Data ~400M Images ~330K Images ~5B Images
Type of Imagery Drawings, Paintings, or Photos Painterly, Aesthetically-Pleasing Images Photorealistic Images or Digital Illustrations
Images Prompt: An astronaut riding a horse in a photorealistic style. Image: Image of an astronaut riding a horse in photorealistic style generated by DALL-E Prompt: /imagine prompt angry cat Image: An angry cat generated by Midjourney Prompt: astronaut looking at a nebula , digital art , trending on artstation , hyperdetailed , matte painting , CGSociety Image: Four different images of an astronaut looking at a nebula as a hyperdetailed digital art generated by Stable Diffusion
Accuracy Accuracy decreases as prompt complexity increases Generally accurate for all images Increased accuracy due to Negative Prompts
Customizability Able to customize specific portions of the image on each iteration of image generation Extensive customizable options Extensive customizable options
Prompt Sizing 400 characters 6,000 characters 320 characters
Uniqueness You can combine multiple images together to create a unique image You can create a prompt by providing an image and then use that prompt to create new imagery The Negative Prompt feature allows you to specify what you do NOT want to achieve your goal

Here’s a concise summary of the AI image generating tools:

Midjourney: A comprehensive and powerful tool that handles extensive prompts, accommodating text and text and image-based input. It is known for generating mostly accurate imagery. Its images look like paintings.

DALL-E: The most user-friendly tool among the three, capable of generating distinctive drawings, paintings, or photos by skillfully combining specific elements from generated imagery to create unique images.

Stable Diffusion: A highly customizable AI tool that allows us to specify what we don’t want, resulting in remarkably accurate image generation. Trained on most of the images of its competitors, it generates four images for its users to choose from. The generated images are photorealistic or digital illustrations.

The choice between these generators often depends on the specific requirements of the user or project. For those seeking highly detailed and imaginative creations, DALL-E might be preferable. Meanwhile, individuals aiming for realistic visualizations might find Midjourney more suitable. Stable Diffusion’s appeal lies in its customization capabilities, making it an excellent choice for those who seek control over image generation.


Stable Diffusion has solidified its position as a frontrunner in the realm of AI-driven image generation, excelling in both text-to-image and image-to-image creation. Its unparalleled expertise in rendering human features, especially faces and hands, is a testament to its prowess in this domain. Trained on an extensive dataset of approximately 5 billion text-to-image examples, Stable Diffusion stands out as one of the most extensively trained AI models for image generation.

In comparison to its competitors, Stable Diffusion emerges prominently in two significant aspects:

  1. Negative Prompts: Offering users the ability to specify what elements to avoid during image generation sets Stable Diffusion apart. This feature minimizes unwanted or unintended aspects, allowing for greater control and precision in the generated images.

  2. Photorealistic Imagery: Among its strengths, Stable Diffusion shines in producing images with a remarkable semblance to reality. The generated visuals often exhibit high levels of realism and authenticity, contributing to its appeal in various applications.

Moreover, what distinguishes Stable Diffusion is its post-generation image modification capabilities. Users can refine, tweak, or adjust the generated images extensively, catering to specific requirements and ensuring a perfectly generated image. Stable Diffusion’s ongoing evolution and its exceptional capabilities in generating realistic, customizable, and contextually relevant images prove it to be a valuable tool in generating incredible images.

If you are interested in reading more about how Generative AI can be applied in your daily life, please check out our AI Catalog of articles!