Unlocking Image Outpainting in SDXL: Stable Diffusion & Auto 1111

Unlocking Image Outpainting in SDXL: Stable Diffusion & Auto 1111

Table of Contents

1. Introduction
2. Understanding the Issue with Out Painting and Seam Lines
3. The Control Net Method
   • 3.1 Selecting the Inpainting Model
   • 3.2 Removing the Area to be Out Painted
   • 3.3 Enabling Control Net and Inpainting
4. The Out Paint Mark 2 Technique
   • 4.1 Using the Out Paint Mark 2 Script
   • 4.2 Adjusting the Width and Sampling Steps
   • 4.3 Disabling Control Net
5. Fixing Seams and Blending with Inpainting
   • 5.1 Sending the Image to Inpainting
   • 5.2 Masking the Seam Area
   • 5.3 Adjusting Mask Blur and Sampling Steps
6. Inpainting the Seam and Removing Watermarks
   • 6.1 Generating the Inpainting
   • 6.2 Masking Out Watermark Areas
   • 6.3 Adjusting Mask Blur and Regenerating
7. Conclusion

Introduction

In this article, we will explore the process of out painting with Stable Diffusion, specifically addressing the issue of seam lines in images generated using SDXL models. Seam lines can be a frustrating problem when trying to Blend the out painted area with the original image. We will revisit two workflows for out painting, starting with the Control Net method and then moving on to the Out Paint Mark 2 technique. Additionally, we will discuss how to fix seam lines and blend images using Inpainting.

Understanding the Issue with Out Painting and Seam Lines

Before diving into the different methods of out painting, it is essential to understand the issue of seam lines. Seam lines occur when the out painted area does not blend seamlessly with the original image. This can result in a visible line where the two areas meet, which can be frustrating for users trying to achieve a seamless blend. In the following sections, we will explore two methods to minimize and eliminate seam lines.

The Control Net Method

The Control Net method is the first approach we will explore for out painting. This method involves using an inpainting model and the Control Net feature to ensure a smooth blend between the out painted area and the original image.

3.1 Selecting the Inpainting Model

To start, it is crucial to choose the right inpainting model for the realistic output. The AO Zia photo inpainting model with VAE B baked into it is recommended for this purpose. Other models can be used Based on preference and experimentation.

3.2 Removing the Area to be Out Painted

Once the inpainting model is selected, it is necessary to remove the area that will be out painted. This can be done by using the original image or part of the prompt. If using the prompt, it is advisable to remove the portion already present in the generated image to focus on what will be added to the outside.

3.3 Enabling Control Net and Inpainting

After removing the unwanted area, enable the Control Net feature and select the inpaint-only option. It is vital to prioritize Control Net and select the image to be resized and filled. Adjust the width and Height Dimensions to add extra pixels to the outer edges. Increasing the sampling steps and selecting the appropriate model will enhance blending. Finally, set the dooy strength to a high value to ensure a successful out painting process.

The Out Paint Mark 2 Technique

The Out Paint Mark 2 technique offers an alternative approach to out painting, which may yield better results in certain scenarios. This method requires fine-tuning specific settings to achieve a seamless blend between the out painted area and surrounding elements.

4.1 Using the Out Paint Mark 2 Script

To start, utilize the Out Paint Mark 2 script, known for its improved performance compared to previous versions. This script handles edges more effectively and can lead to better blending results.

4.2 Adjusting the Width and Sampling Steps

When working with the Out Paint Mark 2 script, it is crucial to adjust the width and sampling steps. Increasing the width can extend the out painted area, providing more space for blending. Adjusting the sampling steps can improve the overall smoothness of the blend.

4.3 Disabling Control Net

Unlike the Control Net method, the Out Paint Mark 2 technique does not require the use of Control Net. The focus here is on using the script effectively and fine-tuning the settings to achieve the desired blend.

Fixing Seams and Blending with Inpainting

In some cases, seam lines may still be present even after using the Control Net or Out Paint Mark 2 method. To address this issue, we will explore how to fix seams and blend the images seamlessly using the Inpainting feature.

5.1 Sending the Image to Inpainting

After completing the out painting process, send the image to the Inpainting tab for further adjustments. This allows for masking specific areas where seams are present and applying inpainting techniques to blend them seamlessly.

5.2 Masking the Seam Area

Create masks over the seam areas using the masking tools available in the AI software. Adjust the mask blur to blur the edges of the masks, and ensure they blend smoothly with the original image.

5.3 Adjusting Mask Blur and Sampling Steps

Fine-tune the mask blur and sampling steps values to achieve optimal blending results. Increasing the mask blur smooths out the transition between the painted and original areas, while adjusting the sampling steps can enhance the overall blend.

Inpainting the Seam and Removing Watermarks

To finalize the seamless blend and remove any watermarks or unwanted elements, we will utilize the Inpainting feature.

6.1 Generating the Inpainting

Send the image to the Inpainting tab to generate the final output. Adjust the settings to match the desired outcome, such as removing watermarks, unwanted objects, or minor imperfections.

6.2 Masking Out Watermark Areas

Mask out the areas where watermarks or other unwanted elements are present. Use masking tools to create accurate and precise masks, ensuring the AI focuses on inpainting those specific regions.

6.3 Adjusting Mask Blur and Regenerating

Fine-tune the mask blur and regenerate the image to effectively remove watermarks and blend them seamlessly with the surrounding elements. Experiment with different settings to achieve the desired result.

Conclusion

In this article, we explored various techniques and methods to achieve seamless blending when out painting with stable diffusion. By utilizing the Control Net method or the Out Paint Mark 2 technique, combined with effective masking and inpainting, users can create visually stunning images without visible seam lines. The key is to experiment with different settings, adjust mask blurs, and fine-tune sampling steps for optimal results. With these techniques, users can overcome the challenges of out painting and achieve a desirable outcome.