Best Practice Workflow for Automatic 1111 – Stable Diffusion

AIKnowledge2Go
26 Jun 202307:59

TLDRThe video presents a detailed workflow for using the stable diffusion model in an automatic setting with a focus on creating semi-realistic renderings. The creator shares tips on setting up the interface, choosing the right model, and optimizing parameters for quality output. They guide the audience through the process of refining an image of a female astronaut and an exploding space station, emphasizing the importance of denoising, sampler selection, and resolution adjustments. The video concludes with a final touch using an upscaler for enhanced detail and shares insights on avoiding common mistakes in digital art creation.

Takeaways

  • 🎨 The video presents a workflow for stable diffusion in automatic 1111 using a semi-realistic model called 'ref animated'.
  • 🔧 To enable the clip skip feature, go to settings, navigate to user interface, and type 'clip stop at last layers' in the Quick Settings list.
  • 🖼️ Euler a is recommended for prompt engineering due to its efficiency, especially when experimenting with different settings.
  • 📏 The recommended dimensions for the model are 768x432, which is a 16:9 resolution suitable for screen wallpapers and avoids deformations.
  • 🚀 Increasing the batch size to 8 allows for the selection of better quality images from the rendered outputs.
  • 🌟 The DPM plus sampler with 2m arrows is used for refining the image, enhancing details and making subtle changes.
  • 🔄 Denoising strength should be set between 0.4 and 0.7, depending on the desired level of change in the final image.
  • 🖌️ Over painting is used to correct errors and add details, such as adding a missing leg to the astronaut character.
  • 👁️ The importance of checking the mask settings in the intent stage is emphasized to ensure the correct elements are retained or modified.
  • 📈 The final step involves upscaling the image using the RS Rugen 4X, anime 6B option for a semi-realistic look.
  • 📚 The video creator plans to release more tutorials, including one about common mistakes in painting hands with automatic 1111.

Q & A

  • What is the primary focus of the video?

    -The primary focus of the video is to demonstrate the best workflow for stable diffusion in automatic 1111, using a semi-realistic model called ref animated, and to provide useful tips, tricks, and insights for creating high-quality renderings.

  • How can the clip skip slider be enabled in the settings?

    -To enable the clip skip slider, go to settings, navigate down to user interface, click into Quick Settings, and enter 'clip stop at last layers'. A restart of the UI is required for the changes to take effect.

  • What is the recommended resolution setting for the Euler a model?

    -The recommended resolution setting for the Euler a model is 768 x 432, which is a 16:9 ratio suitable for most screens, and helps avoid deformations and unwanted visual artifacts.

  • Why is it important to choose the right sampler during the rendering process?

    -Choosing the right sampler is important because it affects the level of detail and the overall quality of the final image. In the workflow, DPM plus with 2m arrows is used to enhance details and introduce some changes without completely altering the image.

  • What is the purpose of denoising strength in the image processing stage?

    -The denoising strength setting is used to control the amount of change introduced to the image during the processing. A value between 0.4 and 0.7 is recommended, with 0.7 allowing for more changes and 0.4 for minimal changes.

  • How does the video creator address the issue of the missing leg in the astronaut's image?

    -The creator plans to use Over Paint to fix the missing leg, ensuring that the astronaut appears complete and realistic in the final rendering.

  • What is the role of the mask in the Over Paint process?

    -The mask plays a crucial role in the Over Paint process by allowing the creator to focus on specific areas of the image, such as the face, without affecting the rest of the rendering.

  • Why is the upscaler RS Rugen 4X Anime 6B a preferred choice for the final image enhancement?

    -The RS Rugen 4X Anime 6B upscaler is preferred because it works well with the semi-realistic look of the ref animated model, enhancing the details and quality of the image without losing the original aesthetic.

  • What are some additional resources available for learning about automatic 1111 and related techniques?

    -The video creator mentions that they have a Patreon page where they share prompts and tutorials, including one about in painting hands and seven common mistakes people make with automatic 1111.

  • How does the video creator engage with the audience and gather feedback?

    -The video creator engages with the audience by reading and responding to comments, using audience feedback to inform the content of future videos and tutorials.

Outlines

00:00

🎨 Best Workflow for Stable Diffusion in Automatic 1111

This paragraph introduces the speaker's personal opinion on the best workflow for using the Stable Diffusion in Automatic 1111, a semi-realistic model known for its beautiful renderings. The speaker shares tips and tricks for using the model, including setting the CLIP skip to 2 and navigating through the settings to enable certain features. A prompt featuring a female astronaut and a space station with an exploding backdrop is prepared and shared on the speaker's Patreon page. The paragraph emphasizes the importance of settings such as Euler a for prompt engineering and the dimensions 768x432 for optimal model handling without deformations. The speaker also discusses the batch size and rendering process, highlighting the selection of appealing images and the decision-making process in choosing the best outcome.

05:02

🖌️ Refining the Image with Sampler and Denoising Settings

In this paragraph, the speaker delves into the process of refining the generated image by altering the sampler to DPM plus 2m arrows and resizing the image for enhanced detail. The denoising strength is set between 0.4 and 0.7, allowing for controlled image changes. The speaker also explains the rationale behind setting the batch count to three for a variety of options. The paragraph further discusses the use of 'send to image to image' for maintaining composition while changing the sampler. The speaker then addresses a leg issue in the image and explains the importance of mask usage in the settings for corrections. The paragraph concludes with the speaker's satisfaction with the final facial details after scaling and denoising adjustments.

Mindmap

Keywords

💡Stable Diffusion

Stable Diffusion is a term used in the context of AI-generated art, referring to a method or process that creates stable, coherent images from textual descriptions. In the video, it is the primary technique being discussed for generating artwork, with the presenter sharing their best workflow for utilizing it effectively.

💡Ref Animated

Ref Animated is mentioned as a semi-realistic model capable of producing beautiful renderings within the context of AI art generation. It is recommended for users who are looking to create visually appealing images, and the video encourages trying it out for its impressive capabilities.

💡CLIP Skip

CLIP Skip is a feature or setting within the context of the AI art generation workflow that the presenter uses to refine the output. It is a technical term related to the process of generating images and is adjusted according to user preferences and the desired outcome.

💡Euler a

Euler a is mentioned as a good choice for prompt engineering in the context of AI-generated art. It is a term related to the algorithms used in the image generation process, and the presenter suggests using it for its speed and efficiency when experimenting with different prompts.

💡Resolution

Resolution refers to the dimensions of the generated images, with specific mention of 768 being the maximum that most models can handle without causing deformations. In the context of the video, resolution is an important aspect of the settings and adjustments made to achieve the desired visual quality.

💡Batch Size

Batch Size is a term related to the number of images generated at once in the AI art creation process. In the video, the presenter increases the batch size to 8 to select from a variety of images and find the best ones that meet the desired criteria.

💡Denoising Strength

Denoising Strength is a parameter used in the AI art generation process to control the level of change or refinement applied to the generated images. It is a value between 0.4 and 0.7, with higher values leading to more changes and lower values resulting in minimal alterations.

💡Image-to-Image

Image-to-Image is a term used in the context of AI art generation to describe a process where an existing image is used as a base to create a new, modified version. This is different from starting with a completely new image, and is used to maintain the composition while making specific adjustments.

💡Intent

Intent, in the context of the video, refers to a feature or process within the AI art generation software that allows users to make manual adjustments or corrections to the generated images. It is used to fix errors or enhance certain aspects of the image.

💡Upscale

Upscale is the process of increasing the resolution of an image, typically to enhance its quality and detail. In the context of the video, it is the final step in the workflow where the generated image is refined for better visual output.

💡Tutorial

A tutorial, as mentioned in the video, is an educational content format designed to teach or guide viewers on a specific topic or skill. The video itself is a tutorial on AI art generation, and the presenter also mentions upcoming tutorials on related subjects.

Highlights

The video presents the best workflow for stable diffusion in automatic 1111.

The recommended model for this workflow is ref animated, which is semi-realistic and capable of beautiful renderings.

CLIP skip is set to 2, a setting that the creator often uses and shares despite not usually revealing it.

To make the CLIP skip slider appear, navigate to settings, user interface, and Quick Settings, then enter 'clip stop at last layers'.

A prepared prompt featuring a female astronaut, a space station, and an exploding station in the background is shared on the creator's Patreon page.

Euler a is preferred for prompt engineering due to its efficiency.

The dimensions 768x432 are chosen to avoid deformations and maintain a 16:9 resolution.

Batch size is increased to 8 for faster selection of quality images.

The process involves rendering images and selecting the most promising ones for further refinement.

Cyrus fix is not used in this workflow as it can significantly alter the composition.

The image-to-image method is used to maintain the composition while changing the sampler.

DPM plus with 2m arrows is chosen as the sampler for its detail enhancement capabilities.

Denoising strength is set between 0.4 and 0.7 to control the level of changes.

Batch count is set to three to provide a range of options and introduce some surprises.

The importance of subscribing is emphasized as it indicates the creation of an audience and helpful content.

An issue with the leg of the astronaut is identified and corrected in the next steps.

The final step involves upscaling the image using the RS Rugen 4X, anime 6B setting for a semi-realistic look.

The creator plans to release more videos, including a tutorial on painting hands and common mistakes with automatic 1111.