Create stunning images with AI-powered In-/Outpainting

Table of Contents

1. Introduction
2. Creating a New Document
3. Adding Selection
4. Generating Images with AI
5. Activating Selection Mode in Photoshop
6. Outpainting
7. Inpainting
8. Using Content-Aware Fill
9. Adding a Castle in the Forest
10. Fixing Glitches with Mid-Journey AI

How to Make In and Out Painting with AI

Are You tired of manually painting in and out parts of your images? Look no further than AI-powered outpainting and inpainting. In this tutorial, we will guide you through the process of using AI to Create stunning images with ease.

1. Introduction

AI-powered image editing has revolutionized the way we approach image manipulation. With the help of AI, we can now generate images, remove unwanted elements, and even add new elements seamlessly. In this tutorial, we will focus on outpainting and inpainting, two techniques that can save you time and effort in your image editing process.

2. Creating a New Document

To get started, create a new document with 1500 to 1500 pixels. Make sure that it is transparent before you start using the plugin. Please add a selection, and you can use the selection button here. Now click on from text and Type in flowers. We Are using the grpc server here in this tutorial. Click on OK, and the AI generates images of flowers.

3. Adding Selection

Click on the info icon to get information about the task. Click on the task itself to open the preview window. Here, I choose the Second one, which looks nice. I activate selection mode in Photoshop now, and the selection is used as an AI generation window. I move it to the right so that it half-overlaps the image. I recommend activating the use all visible layers option in settings. I am using the Eraser to remove the text.

4. Generating Images with AI

Click on outpainting to start out painting now. I select the task, and the window opens again. Now I switch to the layer mode. I browse now through the generated images, select one, and close the preview. You see the flower image has been perfectly extended. I repeat the process now. The lettering from the last picture has also been extended. I remove it on both layers and do another outpaint.

5. Activating Selection Mode in Photoshop

I move now the selection to the next row. If you have dull E2 configuration activated, you can switch to it and Continue outpainting. It uses a larger area for image generation. I finish now the out painting to cover all the canvas. You will find the link to the image for the next example in the description.

6. Outpainting

First, I will try the built-in content-aware fill. Results are not very good. Now I will make a copy of the background layer by dropping it on the plus icon and make it invisible. Now I select the eraser. I prefer to use the Pencil mode because it is needed for Dolly too. For the grpc server, it does not make a difference, but I recommend it anyway. Also, I set it to full hardness. Now I remove the crane in the lower part, which covers the house.

7. Inpainting

I make a square selection now and select also the surroundings. I click on outpainting and type in houses. Looks good now. Before I continue, I merge the two layers and erase the other parts of the crane. I will do the inpainting in two steps. As a prompt, I use just sky. You see that the top of the building in the back has also been restored. You see that you do not need content-aware fill anymore. It also offers different variations for the tip of the church. Nice.

8. Using Content-Aware Fill

In this lesson, I am using an image from unsplash.com. You will find a link to the image in the video description. First, I resize the image to 1500 pixels in width. I extend the canvas to 2500 pixels, and please make sure that the image is left-aligned. I prefer to use the Eraser in pencil mode with full hardness. The grpc server does not care about this, but for Dolly A2, it is required. The meadow does not look very nice in this wedding photo, so I remove the brown stones here.

9. Adding a Castle in the Forest

Now I select the erased part and also something outside of it so that the AI gets some Context. I told the AI that I want to see a meadow with flowers. Often it is needed to play around a bit to find a good matching prompt to the scene. With the info icon, I am getting a quick preview. I am switching to layer preview. I am merging the layers and remove some more parts of the scene. I make another infill now, and again, this time I change the selection a bit. It is impressive how the results are fitting in a very natural way to the scene. It does not look like a fake.

10. Fixing Glitches with Mid-Journey AI

I will Show that the content-aware fill does not do a very good job here as well. I will make an infill with a sky prompt, like in the last lesson. Now I use an outpaint to enlarge the landscape. It is very similar to the first lesson. Make sure to cover enough parts of the image as well. You see that there are nice results here, and I think I will take this one. I checked before and after the AI retouching. Now nice, but I want something more. Now I want to add a castle in the forest in the background.

Sometimes the AI fills only the background even if there is a new object requested. In this case, try to change the selection and prompt. But now we have some nice castles to choose from. Cool. Now I import an image which I generated with the mid-Journey AI. Mid-Journey offers the best quality at the moment AI, so a direct integration of mid-journey in our plugin would be amazing. You see that there are some glitches in this generated image like a bad-looking HAND or face.

We try to fix it now. This is not so easy because AI-generated hands are always difficult, and here we have a similar tone of skin and desert. First, I make a selection of the upper body so the AI has got size reference for a new hand. I remove the old one. I am requesting nine images now because most will not be good. I will try again. It would be also possible to try generating hands in Dalai 2 instead.

As you can see, most hands are not so good, and I realize that the AI does not know that this should belong to a human. I guess the sand confuses it here, so I change the prompt to a female hand and try again. Did I mention already that a direct integration of mid-journey would be nice? So we need a mid-journey application programming interface, short called an API. What do you want for Christmas? I wish Santa brings me a mystery API mid-journey. Oh, where did that bad editing come from? Must have been inhuman.

Third try now. I added photo to the modifier and beautiful to the prompt. Now it looks good, and she is holding some plant. Nice. I switch the face now. Should be easy. Cool. I forgot to Record the video, but I also fixed the leg position. Thanks for watching the tutorial. More to come.

Highlights

• AI-powered outpainting and inpainting can save you time and effort in your image editing process.
• With the help of AI, we can now generate images, remove unwanted elements, and even add new elements seamlessly.
• The grpc server and Dolly A2 are two AI-powered tools that can help you with outpainting and inpainting.
• Mid-Journey AI offers the best quality at the moment AI, and a direct integration of mid-journey in our plugin would be amazing.
• Sometimes the AI fills only the background even if there is a new object requested. In this case, try to change the selection and prompt.

FAQ

Q: What is outpainting? A: Outpainting is the process of extending an image beyond its original boundaries using AI.

Q: What is inpainting? A: Inpainting is the process of restoring missing or damaged parts of an image using AI.

Q: What is content-aware fill? A: Content-aware fill is a built-in feature in Photoshop that uses AI to fill in missing or unwanted parts of an image.

Q: What is mid-Journey AI? A: Mid-Journey AI is an AI-powered tool that offers the best quality at the moment AI for image editing.

Q: Can AI-generated images look natural? A: Yes, AI-generated images can look very natural and seamless when used correctly.