stable diffusion img2img walkthrough : paint an bussin watercolor in minutes (no cap) Krita + Koi

koiboi
7 Sept 202207:33

TLDRThis video tutorial showcases a method to transform simple sketches into sophisticated watercolor paintings using Krita and the Koi plugin. By following a step-by-step process, viewers learn to create a digital canvas, utilize the AI-driven plugin for image enhancement, and experiment with various settings such as sketch strength and style parameters to achieve a range of artistic outcomes. The result is a quick and accessible way to produce aesthetically pleasing artwork, even for those with limited artistic skills.

Takeaways

  • 🎨 The process described is an image-to-image walkthrough using Krita and the Koi plugin to transform simple drawings into polished artwork.
  • 🖌️ Krita is a free software that can be easily installed on any computer and is used to create a new canvas with specific dimensions preferred by the neural network (512x512).
  • 🔧 The Koi plugin needs to be installed by the user and, although a bit tricky, there are videos available to guide through the process.
  • 🖼️ The starting point is a basic drawing, such as a lighthouse, with emphasis on covering the entire canvas to avoid misinterpretation by the AI.
  • 🎨 The AI is instructed to create a watercolor painting with specific qualities and style, such as 'beautiful' and in the style of a renowned watercolor artist like J.M.W. Turner.
  • 🔄 The AI refines the image through multiple iterations (up to 120 steps in the example), with the user adjusting parameters like sketch strength and base seed for different outcomes.
  • 🌟 The 'dream' function of the plugin is used to transform the sketch into a detailed painting, with the potential for a high level of creativity and variation.
  • 📈 Increasing the number of steps and decreasing sketch strength can lead to more refined and aesthetically pleasing results.
  • 🔍 The final images are more impressive when viewed from a distance and can benefit from upscaling to reduce blurriness.
  • 💡 The walkthrough encourages users to experiment with different settings and to use the waiting time for the AI to process the images creatively.
  • 🔗 Additional resources and requests can be found in the description of the video, inviting viewer interaction and feedback.

Q & A

  • What is the main subject of the walkthrough?

    -The main subject of the walkthrough is to demonstrate how to use image-to-image technology with Krita and the Koi plugin to transform a simple drawing into a more complex and aesthetically pleasing artwork, specifically a watercolor painting of a lighthouse.

  • Why is Krita chosen as the software for this walkthrough?

    -Krita is chosen because it is a free and easily installable software on any computer, making it accessible for users to follow along with the walkthrough.

  • What is the recommended canvas size for the neural network in this walkthrough?

    -The recommended canvas size is 512 by 512 pixels, as it is the size that the neural network prefers.

  • What is the role of the Koi plugin in this process?

    -The Koi plugin is used to enhance the initial drawing by applying the image-to-image technology, which transforms the basic sketch into a more refined and visually appealing image.

  • Why is it important to cover the entire canvas with color before using the Koi plugin?

    -It is important to cover the entire canvas because any uncolored pixels may be interpreted as black by the machine learning model. This could lead to unwanted results, as the model may incorporate the black into the final painting in unexpected ways.

  • How many variations does the tutorial ask for in the initial attempt?

    -The tutorial initially asks for four different variants to see how they turn out.

  • What is the significance of the 'steps' parameter in the process?

    -The 'steps' parameter determines how many times the neural network attempts to refine the image, aiming to make it better with each iteration.

  • What does the 'base seed' parameter control in the Koi plugin?

    -The 'base seed' parameter allows for the recreation of the same effect or result if the same seed and parameters are used, ensuring consistency in the output.

  • How does adjusting the 'sketch strength' affect the final image?

    -Adjusting the 'sketch strength' tells the AI how much attention to pay to the original drawing and how much freedom to give it in creating the final image. A lower sketch strength gives the AI more freedom.

  • What is the purpose of making a new layer from all visible elements before clicking 'dream'?

    -Creating a new layer from all visible elements ensures that the AI does not interpret any part of the painting as black, which could lead to a predominantly black and therefore undesirable result.

  • Why might the images look better from a distance?

    -The images might look better from a distance because smaller details that may be blurry or imperfect become less noticeable, allowing the overall composition and color scheme to stand out more.

  • What is suggested for users to do while waiting for the AI to generate the images?

    -The tutorial suggests that users can engage in other activities, such as changing their shirt or pursuing other creative ideas, during the approximately three minutes it takes for the AI to generate the images.

Outlines

00:00

🎨 Transforming Basic Art into Sophisticated Creations with Krita and KOI

The paragraph introduces the concept of transforming simple, unpolished drawings into more sophisticated and visually appealing images using Krita, a free digital painting software, and the KOI plugin. The speaker shares their preference for this method and guides the audience through the process of creating a new canvas with specific dimensions to cater to the neural network's requirements. The importance of covering the entire canvas is emphasized to avoid misinterpretation by the AI. The speaker then provides a step-by-step guide on using the KOI plugin to enhance a basic lighthouse drawing, including the installation of the plugin, the use of specific codes, and the selection of parameters such as the style, number of variations, steps for refinement, base seed, and sketch strength. The goal is to create an aesthetically pleasing oil painting-like image that resembles the style of a renowned watercolor artist, J.M.W. Turner. The paragraph concludes with the speaker's attempt to generate various versions of the enhanced image and their observations on the results, noting that some images have more aesthetic value than others and suggesting that increasing the number of steps could improve the outcome.

05:02

🔍 Enhancing Image Quality and Experimenting with AI Freedom in Art

In this paragraph, the speaker continues the discussion on refining the AI-enhanced image by adjusting specific parameters. They decide to decrease the sketch strength to allow the AI more freedom, aiming for a more detailed and potentially expressionistic result. The speaker keeps the seed the same for comparison and initiates another round of image enhancement. The results show a noticeable improvement in image quality, with some versions appearing more appealing and less blurry. The speaker suggests using an upscaler to further enhance the images and shares their observation that the images look better from a distance. They conclude that the process is worthwhile for the amount of effort投入, and encourage the audience to explore their own creative ideas while waiting for the AI to generate the images. The speaker also mentions making changes to their video appearance based on feedback and invites further suggestions from the audience.

Mindmap

Keywords

💡stable diffusion

Stable diffusion is a term used in the context of image generation and enhancement using artificial intelligence. It refers to a process where an AI model, trained on a large dataset of images, is employed to create or modify images in a way that maintains the stability and coherence of the visual elements. In the video, stable diffusion is used to transform a simple sketch into a detailed watercolor painting of a lighthouse, demonstrating the AI's ability to understand and generate complex visual content.

💡Krita

Krita is a free and open-source digital painting software that provides a platform for artists to create and edit images. It is user-friendly and compatible with various computer systems. In the video, Krita serves as the primary tool for the artist to draw the initial sketch of the lighthouse, which will later be transformed using the stable diffusion process.

💡Koi plugin

The Koi plugin is an add-on for Krita that integrates the capabilities of AI-based image enhancement and generation tools directly into the software. It allows users to utilize features of stable diffusion without the need for complex coding or command-line interfaces. In the video, the Koi plugin is essential for the artist to apply the stable diffusion process to their lighthouse sketch, enabling them to create a detailed watercolor painting with the click of a button.

💡neural network

A neural network is a series of algorithms modeled loosely after the human brain, designed to recognize underlying relationships in a set of data through a process that mimics the way the human brain operates. In the context of the video, the neural network is the AI component that takes the artist's sketch and transforms it into a detailed image through the stable diffusion process, understanding and applying complex visual patterns to generate a realistic watercolor painting.

💡watercolor painting

Watercolor painting is a form of art that uses water-soluble pigments and typically involves the artist applying transparent or semi-transparent washes of color onto paper or other suitable surfaces. In the video, the goal is to create a digital representation of a watercolor painting of a lighthouse, leveraging the stable diffusion process to achieve the visual effects characteristic of this traditional art form.

💡J.M.W. Turner

Joseph Mallord William Turner, commonly known as J.M.W. Turner, was an English Romantic landscape painter, watercolorist, and printmaker. He is known for his expressive use of color and light, and for the way he captured the atmosphere and mood in his works. In the video, the artist references Turner's style to guide the AI in creating a watercolor painting that emulates the characteristics of Turner's renowned artwork.

💡sketch strength

Sketch strength is a parameter used in the stable diffusion process that determines how much emphasis the AI places on the original sketch provided by the user. A lower sketch strength value gives the AI more freedom to deviate from the sketch and create a more detailed or expressive image, while a higher value would make the AI adhere more closely to the original sketch. In the video, adjusting sketch strength is a key step in fine-tuning the AI's output to achieve the desired level of detail and artistic style.

💡base seed

The base seed in the context of AI-generated images is a starting point or a set of initial values that the neural network uses to generate an image. The same base seed, when used with the same parameters, will produce consistent results, allowing for the replication of specific visual outputs. In the video, the artist provides a base seed value to ensure that viewers can recreate the exact effects and outcomes if they wish to.

💡iterations

Iterations in the context of the video refer to the repeated application of the stable diffusion process to refine and improve the AI-generated image. Each iteration represents a cycle of enhancement, where the neural network adjusts and optimizes the image based on the set parameters. The artist in the video conducts multiple iterations to achieve the best possible result, experimenting with different settings to find the optimal balance between the original sketch and the AI's creative enhancements.

💡upscale

Upscaling an image involves increasing its resolution while maintaining or improving its visual quality. This process is often used to make low-resolution images appear clearer and more detailed when viewed at larger sizes. In the video, the artist suggests using an upscaler to improve the clarity of the AI-generated watercolor paintings, which may appear blurry at their original resolution.

💡aesthetic value

Aesthetic value refers to the appreciation of beauty or good taste in art, which is often subjective and can vary from person to person. In the video, the artist discusses the aesthetic value of the AI-generated images, noting that some of the results have a certain charm or appeal, despite not being perfect. The artist's goal is to create images that are not only technically impressive but also visually pleasing and engaging.

Highlights

Stable diffusion img2img technique allows for transforming simple sketches into sophisticated artwork.

The process has gained significant popularity online due to its ability to produce high-quality images quickly.

Krita, a free software, is used as the primary tool for this walkthrough.

A canvas size of 512 by 512 is recommended for optimal neural network performance.

The Koi plugin is essential for the img2img process and can be installed with some effort.

Starting with a basic subject, like a lighthouse, provides a clear direction for the AI to follow.

Covering the entire canvas with a base color prevents unwanted black pixels from influencing the AI.

Adding artistic style and detail qualifiers can guide the AI to produce a specific type of artwork.

JMW Turner's style is suggested for its beauty and effectiveness in watercolor paintings.

Requesting multiple variants can lead to a range of different outcomes, increasing the chances of satisfaction.

Adjusting the number of steps and sketch strength allows for fine-tuning the final image.

The AI's output can be significantly improved by giving it more freedom and increasing the steps.

Upscaling the final images can reduce blurriness and enhance the visual quality.

Viewing the images from a distance can reveal their aesthetic value more effectively.

The process is practical and can yield impressive results with as little as 20 minutes of work.

Experimenting with different settings and iterations can lead to unique and creative outcomes.

The walkthrough provides a step-by-step guide for users to recreate the process and achieve similar results.