(KRITA AI) STEP-BY-STEP Using Stable Diffusion in Krita

Intelligent Image
3 Jun 202419:51

TLDRIn this tutorial, the creator demonstrates using the Stable Diffusion plugin in Krita to enhance a sketch. Starting with a 1000x1000 pixel canvas, they employ various techniques such as control nets and live mode to refine details like hands and hair. They also experiment with upscaling and style transfer, ultimately achieving a more polished artwork. The video offers insights into leveraging AI for artistic refinement in Krita.

Takeaways

  • 🎨 Start with a small canvas size (e.g., 1000x1000 pixels) for faster generation when using Stable Diffusion in Krita.
  • 🖌️ Use the 'vxp turbo checkpoint' for style-specific prompts to maintain consistency in image generation.
  • 📝 Keep prompts simple and descriptive, focusing on the elements present in the image for better results.
  • 🔍 Use control nets like 'line art' to guide the AI in refining specific parts of the image, such as hands or facial features.
  • ⚙️ Adjust control net strength and range to balance the influence of the original sketch and the AI's creative liberties.
  • 🔄 Experiment with different control net settings and batch sizes to find the best balance between detail preservation and artistic enhancement.
  • 🖥️ Consider upscaling the image using a model that complements the original style, but be cautious of detail loss and image degradation.
  • 🎭 Use style transfer with control nets to apply the style of one image to another, adjusting strength to maintain the desired look.
  • 🖼️ After AI processing, manually refine the image by adding or removing details, and use layer masks to blend changes seamlessly.
  • 🔍 For final touches, use Krita's filters and adjustments to enhance contrast and sharpness, adjusting layer opacity for a balanced effect.

Q & A

  • What is the recommended starting canvas size when using the Stable Diffusion plugin in Krita?

    -The recommended starting canvas size is 1,000 by 1,000 pixels. However, for those with slower hardware, it's suggested to start even smaller, such as 768 x 768 pixels.

  • Why should one start with a smaller canvas size when using the Stable Diffusion plugin?

    -Starting with a smaller canvas size is recommended because the larger the canvas, the slower the generation process will be. Smaller sizes allow for quicker iterations, which is beneficial when starting out and making adjustments.

  • What is a control net and how does it function within the Stable Diffusion plugin?

    -A control net in the Stable Diffusion plugin is a feature that allows users to guide the AI's output to better match certain aspects of an image or style. It can be used to refine specific elements of the generated artwork.

  • How does the 'vxp turbo checkpoint' mentioned in the script influence the style of the generated image?

    -The 'vxp turbo checkpoint' is an sdxl model that affects the style of the generated image. It's used to maintain a consistent style throughout the image generation process, as per the artist's preference.

  • What is the purpose of the 'style prompt' in the Stable Diffusion plugin workflow?

    -The 'style prompt' is used to list out the elements that are in the image and to guide the AI towards creating an image with those specific characteristics. It helps the AI understand what the artist is aiming to create.

  • Why might the artist choose to add 'space and stars' to their prompt in the Stable Diffusion plugin?

    -Adding 'space and stars' to the prompt helps to maintain a Sci-Fi Space theme in the artwork. It guides the AI to include these thematic elements in the generated image.

  • What is the significance of the 'negative prompt' in the Stable Diffusion plugin settings?

    -The 'negative prompt' is used to specify elements or styles that the artist wants to avoid in the generated image. It helps the AI to exclude certain aspects that are not desired in the final artwork.

  • How does adjusting the 'strength' of the control net affect the image generation in the Stable Diffusion plugin?

    -Adjusting the 'strength' of the control net influences how much the AI adheres to the guidance provided by the control net. A higher strength means the AI will more closely follow the control net's guidance, while a lower strength allows for more variation.

  • What is the 'batch size' setting in the Stable Diffusion plugin and how does it impact the image generation process?

    -The 'batch size' setting determines the number of images the AI will generate at once. A higher batch size means the AI will produce more variations for the artist to choose from, but it also increases the processing time.

  • What is the 'live mode' in the Stable Diffusion plugin and how is it used in the workflow?

    -The 'live mode' in the Stable Diffusion plugin allows the artist to make real-time adjustments to the image by interacting directly with the AI. It can be used to clean up or modify specific parts of the generated image without affecting the entire canvas.

  • Why might the artist choose to upscale their image using the Stable Diffusion plugin?

    -Upscaling an image using the Stable Diffusion plugin can help to increase the resolution and detail of the artwork. It's a useful step when the artist is satisfied with the composition but wants to enhance the clarity and quality of the image.

Outlines

00:00

🎨 Introduction to Creating Art with Stable Diffusion

The speaker begins by introducing the process of creating art using the stable diffusion plugin for CR, starting with a blank canvas set at 1000x1000 pixels. They discuss the importance of starting with a smaller canvas for faster generation, especially for those with slower hardware. The speaker then considers different starting points, such as live painting or generating something new. They decide to import a recent sketch to experiment with the vxp turbo checkpoint, setting up a prompt that describes the elements in the image. The aim is to see if the plugin can improve upon the original sketch, focusing on aspects like the hands and adding stylistic elements like 'space' and 'stars' to maintain a Sci-Fi theme.

05:01

🖌️ Refining Art with Control Nets and Live Mode

The speaker experiments with a control net to enhance the sketch, adjusting the strength and range to improve the details, particularly the hands. They encounter issues with the control net creating a neon effect when the strength is too high, so they adjust the settings accordingly. The process involves generating multiple images to select the best outcome. The speaker then switches to live mode to manually clean up areas like the hands and hair, using the control net to maintain consistency. They also discuss the use of transparency masks to revert unwanted changes and the decision to upscale the image for more detail, experimenting with different models and settings.

10:01

🔍 Advanced Techniques for Image Refinement

The speaker explores advanced techniques for refining the upscaled image, discussing the challenges of maintaining detail and avoiding blurriness. They experiment with different control net strengths and denoising levels to improve the image quality. The speaker also attempts to transfer the style from an old piece of art onto the current image, using a style control net and adjusting the range to minimize unwanted object transfers. They find that the style transfer is influenced by the layers selected for control, leading to a discovery about the importance of layer selection in style transfer. The process involves trial and error, with the speaker making adjustments to achieve a more expressive and detailed result.

15:04

🖋️ Final Touches and Conclusion

In the final stages, the speaker focuses on enhancing the facial details, using control layers and refining the image to achieve a more expressive look. They encounter issues with the style layer referencing the wrong image, which is resolved by duplicating and renaming the layer. The speaker then applies the refined image, adjusts the opacity, and uses a transparency mask to blend it with the original. Additional adjustments include adding contrast and sharpness using filter layers. The speaker concludes by expressing satisfaction with the final result, mentioning the possibility of using the AI-generated image as a reference for a manual finish in the future. They also invite viewers to explore more about generative AI plugins in their other videos.

Mindmap

Keywords

Stable Diffusion

Stable Diffusion is a type of deep learning model used in the field of generative AI. It is capable of generating images from textual descriptions. In the context of the video, the presenter uses the Stable Diffusion plugin for Krita, an open-source drawing software, to create images based on prompts and existing sketches. The process involves iterative refinement, where the AI generates images that the artist can then modify to achieve the desired outcome.

Krita

Krita is a free and open-source painting and illustration program. It is designed for digital artists and offers a wide array of tools for creating digital art. In the video, the presenter uses Krita as the platform to integrate the Stable Diffusion plugin, allowing for AI-assisted image creation and manipulation.

vxp turbo checkpoint

In the video, 'vxp turbo checkpoint' refers to a specific model or preset within the Stable Diffusion plugin that is used to generate images in a certain style. The presenter mentions using this checkpoint to maintain consistency with their desired artistic style throughout the image generation process.

Prompt

A 'prompt' in the context of AI image generation is a textual description that guides the AI in creating an image. It typically includes key elements that the user wants to see in the generated image. In the video, the presenter uses a simple prompt like 'girl leather jacket gloves white hair' to describe the main subject of the image they want to create.

Control Net

A 'Control Net' is a feature within the Stable Diffusion plugin that allows the user to guide the AI's image generation process by providing additional reference images. In the video, the presenter uses a line art Control Net to help the AI maintain certain visual elements from the original sketch during the generation process.

Live Mode

Live Mode is a feature in the Stable Diffusion plugin that allows the user to interactively edit an image in real-time, with the AI making adjustments based on the input. The presenter uses Live Mode to clean up parts of the image, such as the hands, by adding or removing details as needed.

Batch Size

Batch Size refers to the number of images the AI will generate at one time based on the given parameters. In the video, the presenter sets the batch size to 10, which means the AI will generate 10 different images for the user to choose from, allowing for selection of the most preferred result.

Upscale

Upscaling in digital art refers to increasing the resolution or size of an image while maintaining or improving its quality. In the video, the presenter upscales the image to 1500 by 1500 pixels using a specific model within the Stable Diffusion plugin to enhance the details and overall quality of the image.

Denoising

Denoising is the process of reducing noise or artifacts in an image to improve its clarity and quality. In the context of the video, the presenter adjusts the denoising strength when using the Stable Diffusion plugin to refine the upscaled image and reduce any blurriness or unwanted visual noise.

Style Transfer

Style Transfer is a technique in AI image generation where the AI adopts the visual style of one image and applies it to another. In the video, the presenter experiments with transferring the style of an older artwork onto the new image to achieve a consistent artistic theme across different pieces.

Highlights

Introduction to using Stable Diffusion in Krita for image creation.

Advising to start with a small canvas size for faster generation.

Recommendation to scale down the canvas for slower hardware.

Using a sketch as a starting point for the image generation process.

Explanation of using the vxp turbo checkpoint for style.

Suggestion to keep the prompt simple when starting with an image.

Importance of adjusting settings to affect the style in the image.

Adding a control net to influence the image generation.

Adjusting control net strength and range for better results.

Using batch generation to create multiple options for selection.

Applying the generated image and discussing areas for manual cleanup.

Switching to live mode for manual adjustments and the use of control nets.

Demonstrating the process of refining gloves and hair in live mode.

Using transparency masks to remove unwanted changes.

Discussing the challenges of upscaling with the vxp Turbo model.

Experimenting with different control nets for style transfer.

Using a control net to maintain details during resampling.

Incorporating an old piece of art to influence the style of the generated image.

Final touches and adjustments to the image using filters and manual cleanup.

Conclusion and satisfaction with the final result of the image.

Mention of future plans to use AI as a reference for manual finishing of artwork.