Stable Diffusion 올바로 사용하기 #1 - 프롬프트와 세팅 설정

DigiClau (디지클로) Lab
26 Mar 202321:25

TLDRThe video script introduces the viewer to the popular feature of Stable Diffusion for generating images from text prompts. It explains the process of using the Stable Diffusion web UI, selecting models like Checkpoint models, and adjusting various settings for image generation. The video also discusses finding and using models from websites like and emphasizes the importance of negative prompts and embedding files to refine the output. The content is informative, offering insights into creating high-quality images and avoiding common pitfalls, all while encouraging viewers to experiment with different settings and models.


  • 📝 The primary function of Stable Diffusion is text-to-image conversion, enabling AI to generate images based on text prompts.
  • 🔧 Various settings options are available in Stable Diffusion to fine-tune the image generation process.
  • 🌐 Stable Diffusion operates through a web UI, making it accessible via a web browser.
  • 🧠 Model checkpoints (or weights) are crucial for image creation, with Stable Diffusion providing a default 1.5 version model for starters.
  • 👨‍💻 A specific example of a text prompt includes creating a high-quality, realistic photo of a woman with brown hair and eyes, detailing the importance of negative prompts to avoid low-quality results.
  • 📈 CB ai ( is recommended for finding a wide range of high-quality checkpoint models beyond the default offerings.
  • 📦 Downloading and implementing alternative models, like the Kyuran Mix, enhances the diversity and quality of generated images.
  • 🛠 The LORA model and texture inversion files, such as Easy Negative, provide fine adjustments to influence specific aspects of the generated image.
  • 📸 Experimenting with various settings, including sampling methods and steps, can significantly affect the output's quality and accuracy.
  • 💬 Negative prompts and embeddings are essential for excluding undesired elements and refining the final image.

Q & A

  • What is the primary function of the Stable Diffusion feature discussed in the script?

    -The primary function of the Stable Diffusion feature discussed in the script is to generate AI images based on text prompts provided by the user.

  • What are the Stable Diffusion Checkpoints and how do they relate to image generation?

    -Stable Diffusion Checkpoints, also known as models, are essential components in the Stable Diffusion process for generating images. They define the base model used to create the images and can be switched to produce different styles or qualities of images.

  • How can users find and use different models for Stable Diffusion?

    -Users can find different models for Stable Diffusion on websites like, which hosts a variety of models with different qualities and characteristics. Users can download the desired model and place it in the appropriate folder within the Stable Diffusion UI directory to use it for image generation.

  • What is the role of the 'Text to Image' option in the Stable Diffusion UI?

    -The 'Text to Image' option in the Stable Diffusion UI is where users input their text prompts to instruct the AI to generate images based on the provided descriptions.

  • What are some of the settings available in the Stable Diffusion UI for users to customize their image generation?

    -The Stable Diffusion UI offers a variety of settings for users to customize their image generation. These include sampling methods, sampling steps, restore face option, tiling, file resolution fixes, batch count and size, CF scale, and seed value.

  • How does the 'Negative Prompt' work in the Stable Diffusion process?

    -The 'Negative Prompt' is used to specify elements or characteristics that should be excluded from the generated image. It helps guide the AI to avoid producing unwanted features in the final image.

  • What is an 'Embedding' in the context of the Stable Diffusion script?

    -In the context of the Stable Diffusion script, an 'Embedding' is a type of file that is used to improve the quality of the generated images by adding specific negative prompts to the text prompt. It helps the AI avoid generating certain undesirable features.

  • How does the 'CFG Scale' setting affect the image generation in Stable Diffusion?

    -The 'CFG Scale' setting adjusts the influence of the text prompt on the generated image. A higher CFG Scale means the AI will more closely follow the text prompt, while a lower scale allows for more AI interpretation and creativity in the image generation.

  • What is the purpose of the 'Seed' value in Stable Diffusion image generation?

    -The 'Seed' value is a unique identifier that determines the randomness of the image generation. Using the same seed value will produce similar images, allowing users to create a series of images with consistent themes or styles.

  • How can users ensure the quality and desired characteristics of the generated images?

    -Users can ensure the quality and desired characteristics of the generated images by carefully selecting and adjusting the settings in the Stable Diffusion UI, such as the model, sampling method, steps, and using embeddings and negative prompts to guide the AI generation process.

  • What are some tips for users to get the best results from the Stable Diffusion image generation?

    -Some tips for getting the best results include experimenting with different models, adjusting the settings according to the desired image quality, using embeddings and negative prompts effectively, and testing various text prompts to find the optimal combination for the intended image.



🖌️ Introduction to Stable Diffusion Image Generation

This paragraph introduces the concept of using Stable Diffusion for text-to-image generation. It explains that the AI creates images based on text prompts and that there are many options to set for the generation process. The video aims to guide viewers on how to use prompts and various settings within the UI of Stable Diffusion. It also encourages viewers to watch the video to learn important aspects of Stable Diffusion and mentions the need to install the software if not already done.


🛠️ Exploring Stable Diffusion's UI and Settings

The paragraph discusses the user interface of Stable Diffusion and how to navigate it. It highlights the importance of selecting the right model for image generation and introduces the concept of 'checkpoint' models. The paragraph also mentions a website called as a resource for finding various models and provides a detailed process of downloading and using a specific model called 'Chillout Mix' to enhance the image generation process.


🎨 Customizing Image Generation with Specific Settings

This section delves into the specifics of image generation using Stable Diffusion. It explains how to use different settings such as sampling methods, steps, and seed values to create unique images. The paragraph also discusses the role of negative prompts and the importance of understanding the model's behavior when generating images based on the provided prompts.


🌟 Enhancing Image Quality with Additional Models

The paragraph focuses on enhancing the quality of generated images by using additional models like 'Lora' to introduce minor variations to existing checkpoint models. It explains the process of downloading and integrating Lora models into the Stable Diffusion workflow and how they can be used to fine-tune the final output. The paragraph also touches on the use of negative prompts and embedding files to further refine the image generation process.


📸 Putting It All Together: Creating Diverse Images

This paragraph concludes the video script by demonstrating how to use various settings, prompts, and models to create a diverse range of images. It encourages viewers to experiment with different prompts and settings to achieve the desired image outcomes. The paragraph also mentions the importance of creativity and imagination in using Stable Diffusion to generate unique and high-quality images.



💡Stable Diffusion

Stable Diffusion is a type of AI model used for generating images from text prompts. It is the main focus of the video, where the user is introduced to its capabilities and various settings. The video explains how to use the Stable Diffusion model to create images that match the text descriptions provided by the user.


Text-to-Image refers to the process of converting textual descriptions into visual images using AI technology. In the context of the video, this is the primary function of Stable Diffusion, where users input text prompts to receive corresponding images. This feature is highlighted as the most commonly used in Stable Diffusion.


Checkpoints in the context of the video are specific versions of AI models used within the Stable Diffusion system. They are crucial for the image generation process, as they determine the style and quality of the output images. The video instructs viewers on how to select and use different checkpoints, such as the 'Turbo AI' and 'Craiyon Mix' models.

💡UI Settings

UI Settings refer to the user interface options and configurations available within the Stable Diffusion web UI. These settings allow users to customize the image generation process, such as adjusting the quality, sampling method, and other parameters to achieve desired results.

💡Negative Prompts

Negative Prompts are instructions given to the AI to avoid certain elements or characteristics in the generated images. They are used to refine the output by specifying what should not be included, helping to prevent unwanted features or mistakes.


Embeddings are AI-trained files that assist in improving specific aspects of the generated images by focusing on particular features or elements. They are used within the text prompts to enhance the quality and accuracy of the AI's output.

💡Craiyon Mix Model

The Craiyon Mix Model is a specific checkpoint in the Stable Diffusion system that is known for generating high-quality, realistic human portraits. It is highlighted in the video as a model that is particularly good for creating images of people.

💡Lora Model

The Lora Model is a smaller, lightweight AI model designed to make minor adjustments to the base checkpoint models in Stable Diffusion. Despite their smaller size, Lora Models can effectively introduce variations and fine-tune the output images according to the user's needs.

💡Sampling Method

The Sampling Method refers to the technique used by the AI to select and combine elements from the model's training data to create the final image. Different sampling methods can affect the quality and style of the generated images, with some being more suitable for certain types of images than others.


Steps in the context of the Stable Diffusion UI refers to the number of iterations or stages the AI goes through to generate an image. Increasing the number of steps can improve the quality and detail of the image, but it also increases the time required for generation.

💡CFG Scale

CFG Scale, or Control Flow Graph Scale, is a parameter in the Stable Diffusion UI that determines how closely the generated image adheres to the text prompt provided by the user. A higher CFG Scale means the AI will follow the prompt more strictly, while a lower value allows for more creative freedom from the AI.


The Seed value is a unique identifier used in the image generation process to ensure consistency and reproducibility. By using the same seed value, users can generate a series of images that are similar to each other, which can be useful for creating cohesive sets of visual content.


Introduction to the Stable Diffusion feature in text-to-image generation.

Explanation of various options and settings available in the Stable Diffusion UI.

Use of the Checkpoint models, also known as 'Stable Diffusion Checkpoints', in image generation.

The importance of choosing the right model for desired image outcomes.

How to find and use different Checkpoint models from websites like

The process of downloading and implementing a new Checkpoint model into the Stable Diffusion UI.

Utilization of the 'Turron Mix' model for creating high-quality, realistic images.

Explanation of the 'Negative Prompts' to guide the AI away from undesired image features.

The role of 'Sampling Method' in determining the quality and style of generated images.

Adjusting 'Steps' for refining image quality and the trade-off with processing time.

The 'Restore Face' option to correct facial features in generated images.

Understanding the 'Tiling' option for creating images suitable for tiling.

The 'File Resolution Fix' feature for upscaling images after generation.

Setting 'UI Scale' and 'Batch Count' for controlling the number of images generated.

Exploring the 'CF Scale' to balance the adherence to the text prompt.

The significance of the 'Seed' option for creating similar images in a series.

Creating an image using the 'Turron Mix' model and various settings.

The introduction and application of 'Lora' models for fine-tuning image features.

Utilizing 'Embeddings' like 'Lora' and 'Negative Embeddings' for enhancing image quality.

The impact of different 'Embeddings' on the final image outcome.

Creating diverse images using a combination of text prompts, models, and settings.

Encouragement for viewers to experiment with different text prompts and settings for unique image creation.