HOW TO MAKE BEAUTIFUL STABLE DIFFUSION IMAGES | Negative Prompts

Binks
13 Jan 202305:02

TLDRIn this video, Binks discusses the significance of negative prompts in the process of generating images with Stable Diffusion. Binks uses ProtoGen version 3.4, a model known for its photorealism, and provides a link to it in the description. The video demonstrates how to use a negative prompt to refine the image generation process, avoiding unwanted elements such as canvas frames and disfigurements. Binks also shares his preferred sampling method, DPM++ SDE Keras, and suggests adjusting sampling steps and height for better portrait results. By incorporating a detailed negative prompt and tailored sampling settings, Binks showcases how to generate a stunning image that aligns with the desired outcome. He encourages viewers to experiment with the AI, learn its workings, and iteratively improve their prompts for better results.

Takeaways

  • 🎨 **Importance of Negative Prompts**: Negative prompts are crucial in guiding Stable Diffusion to avoid unwanted elements in generated images.
  • 🖼️ **Avoiding Unwanted Features**: By including 'canvas frame' in the negative prompt, the AI is instructed to not include a canvas frame in the image.
  • 🚫 **Excluding Undesired Elements**: Negative prompts can prevent the AI from generating disfigured subjects, extra limbs, or mutations.
  • 👩‍🦰 **Customizing the Image**: Tailoring the prompt with specific descriptions and keywords helps in generating a more desired outcome.
  • 🔍 **Detailing the Prompt**: Using a detailed prompt can lead to more accurate and stylized results, as demonstrated by the 'medieval model shoot style' example.
  • 📈 **Sampling Method and Steps**: The choice of sampling method (DPM++ SDE Keras) and the number of sampling steps (e.g., 30) can significantly affect the image outcome.
  • 🖥️ **Image Resolution**: Adjusting the height of the image (from 512 to 768) can provide a more portrait-oriented look.
  • 🔧 **Config Scale Adjustment**: Increasing the config scale (e.g., to 10) can improve the image quality for certain models.
  • 🧩 **Iterative Process**: The process of generating images with Stable Diffusion is iterative, requiring multiple attempts and adjustments.
  • 🎭 **Artistry in AI**: There's an element of artistry involved in understanding how the AI system works and crafting prompts to achieve desired results.
  • 📚 **Learning and Experimentation**: The speaker encourages viewers to learn through repeated experimentation and to refine their prompts and settings.

Q & A

  • What is the main topic of the video?

    -The main topic of the video is discussing the importance of negative prompts in stable diffusion for generating images and addressing why the generated images might not match expectations.

  • Who is the speaker in the video?

    -The speaker in the video is Binks.

  • What version of ProtoGen is Binks using in the video?

    -Binks is using ProtoGen version 3.4, specifically the photorealism release.

  • What is the purpose of using a negative prompt in stable diffusion?

    -The purpose of using a negative prompt is to guide the AI to avoid including certain elements or characteristics in the generated image that the user does not want.

  • What is the default sampling method Binks prefers for ProtoGen?

    -Binks prefers the DPM++ SDE Keras sampling method for ProtoGen.

  • How many sampling steps does Binks usually set for generating images?

    -Binks usually sets the sampling steps to around 30.

  • What is the significance of the negative prompt in the context of generating a portrait of a blonde woman in a medieval style?

    -The negative prompt helps to ensure that the generated image does not include unwanted elements such as a canvas frame or disfigured subject, allowing for a more accurate representation of the desired medieval portrait.

  • How does Binks tailor the original prompt to get closer to the desired image?

    -Binks tailors the original prompt by adding more detailed descriptions and keywords that describe the subject in various ways, which helps the AI generate a more accurate image.

  • What does Binks suggest for users to do if they want to improve their generated images?

    -Binks suggests that users should experiment by clicking the generate button repeatedly, using different prompts and settings to find the best results.

  • Why does Binks recommend checking the 'restore faces' option?

    -Checking the 'restore faces' option helps to ensure that the generated images have more accurate and recognizable human faces.

  • What is the role of the 'config scale' in the image generation process?

    -The 'config scale' is a parameter that Binks adjusts to influence the quality and detail of the generated image; in the video, it is set to around 10 for the specific model used.

  • How does Binks describe the process of working with AI to generate images?

    -Binks describes the process as involving a bit of work and artistry, where users need to understand how the system works and craft their prompts carefully to achieve the desired results.

Outlines

00:00

🎨 Introduction to Negative Prompts in Stable Diffusion

Binks introduces the video by discussing the significance of negative prompts in image generation using Stable Diffusion. He mentions using ProtoGen version 3.4, specifically the photorealism release by Darkstorm 2150, and provides a link to it in the description. Binks also shares his preferred sampling method for ProtoGen, DPM++ SDE Keras, and explains his typical settings for generating images, including sampling steps, height, and config scale. He emphasizes the importance of restoring faces and demonstrates the process of generating an image without initially using a negative prompt, then refining the process by including a detailed negative prompt to achieve a more desired result.

Mindmap

Keywords

💡Negative Prompts

Negative prompts are instructions given to an AI image generation system to avoid including certain elements in the generated image. In the video, Binks discusses the importance of negative prompts in achieving the desired outcome with AI-generated images, particularly when using Stable Diffusion. An example from the script is using 'canvas frame' as a negative prompt to prevent the AI from adding a frame around the generated portrait.

💡Stable Diffusion

Stable Diffusion is an AI model used for generating images from textual descriptions. It is mentioned in the video as the platform where the negative prompts are applied. Binks uses Stable Diffusion to create an image of a 'blonde woman' in a 'medieval' style, adjusting the prompts and settings to refine the output.

💡Protogen 3.4

Protogen 3.4 refers to a specific version of an AI model used for generating photorealistic images. Binks mentions using this version for its capabilities and provides a link to it in the video description. It is used as a basis for generating images in the tutorial.

💡Sampling Method

The sampling method is a technique used in AI image generation to determine how the AI interprets the input prompts and creates the image. Binks prefers the 'DPM++ SDE Keras' method for Protogen, which is mentioned as his favorite for achieving a certain look in the generated images.

💡Sampling Steps

Sampling steps refer to the number of iterations the AI goes through to generate an image. Binks increases the sampling steps to around 30 for better image quality, indicating that more steps can lead to a more refined and detailed output.

💡Config Scale

Config scale is a setting that adjusts the intensity or the 'strength' of the image generation process. Binks increases the config scale to 10 for the model he is using, suggesting that this setting can affect the final appearance of the generated image.

💡Restore Faces

Restore Faces is an option that can be toggled during the image generation process to improve the quality and detail of faces in the generated images. Binks checks this option as part of his default settings when generating images with Protogen.

💡Medieval

The term 'medieval' refers to the Middle Ages period in history. In the context of the video, Binks uses 'medieval' as a style descriptor in the image prompt to guide the AI towards generating an image with a medieval aesthetic.

💡Model Shoot Style

Model shoot style implies a professional and artistic approach to photographing a model. Binks tailors the prompt to include 'model shoot style' to direct the AI towards creating a more stylized and high-quality image.

💡AI Artistry

AI Artistry refers to the creative process of using AI to generate art. Binks emphasizes that there is an element of artistry involved in working with AI image generation, as it requires understanding the system and crafting prompts to achieve the desired result.

💡Tailored Prompt

A tailored prompt is a carefully crafted input given to an AI system to guide it towards a specific outcome. Binks discusses the importance of creating a detailed and specific prompt to direct the AI in generating an image that closely matches the user's vision.

Highlights

The importance of negative prompts in stable diffusion for generating desired images.

Using Protogen version 3.4 for photorealism in image generation.

The option to explore different models on the Civic AI page.

Negative prompts prevent unwanted elements like canvas frames and disfigurements.

Crafting negative prompts to refine the AI's output.

Using DPM++ SDE Keras as the preferred sampling method for Protogen.

Increasing sampling steps to around 30 for better image quality.

Adjusting the height parameter to 768 for a more portrait-oriented look.

Configuring the scale to 10 for optimal results with the specific model.

Restoring faces as a default option to improve the quality of facial features.

The impact of a tailored prompt and negative prompt on the final image.

Describing the subject in multiple ways with various keywords for a detailed result.

The process of running a well-trained model with tailored settings.

The artistry involved in understanding the AI system for better image generation.

Encouragement to experiment with the generate button for different results.

The value of asking questions and engaging with the community for further learning.

The satisfaction of generating stunning images through careful prompt crafting and model training.