Working with poses

Welcome to the first practice lesson! In this lesson, we will talk about working with poses and explore their importance in creating unique images using Artificial Intelligence (AI).

Why do we need poses ?

With the help of correctly chosen pose you can set the mood of the whole image and cope with possible errors of the model. In particular, the problem of generation of fingers on hands, and in general the problem of generation and display of hands.

Understanding pos

Let's start with a general understanding of poses. Poses define the position of the object in the image and can be diverse:

• Hands: Hands can be placed in many different ways: on your face, on your hip, in your pocket, above your head, and so on.
• Gaze: The gaze of a person in an image can point to different objects or directions.
• Position: Sitting, standing, lying down.

Practice

And here are the poses we will be working with. I tried to group everything in such a way that the lesson would be as concise as possible.

Hand positions:

• hand on face,
• hand on hip,
• hand in pocket,
• hand over head,
• arms crossed,
• finger on lip

Work with gaze (universal parameter, can be used not only for working with poses):

• looking over back/shoulder,
• looking into distance,
• looking to at the viewer

For sitting position:

• sit with arched back,
• sit with legs crossed,
• sit with hand on chin

For demonstration I will use this prompt (separately marked what was changed)

For self-generation you can use this link here

Best results

To solidify the understanding of this lesson, I will show you a few of the most versatile poses, for entry level, that will give you the most beautiful results.

The first one is "hand in pocket" - as we can see there are no problems with hands. And all because they are not visible and they are in the pocket. So we don't spoil the picture and we have a good request..

The second query that shows itself very well is "sitting with hand on top of head". Here the reason is the same. The problem areas of the hands and, but hidden from the user's view.

Also a query containing "looking at viewer" shows itself very well. But such a variant as I got is quite rare. This position, if you want to get the same variant as mine, should be combined with the display parameter "close up". But I will tell you more about it in the next lesson.

And for the sake of example, I will show the request in detail, where the problem areas of the model used were clearly visible.

Results with display errors

When creating images, you need to keep in mind where a person will look first when they see the image. In this case, the "hand on face" pose was used, where the eye immediately falls on the crooked hand near the face. This is immediately visible and does not look good.

In turn, for example, when using the pose of one "finger on lip", it is not immediately possible to see another (third) hand. Which was so kindly added to us by the AI thinking that it is necessary to make more crossed arms under the chest. But agree, it is not immediately obvious, because we emphasize that the hand should be near the face and one of the fingers on the lip. Unfortunately, the AI did not reach the lip, while the finger looks in the right direction. This variant is not the worst, and if necessary it can be improved.

Homework

Your homework will be to try different parameters for the poses yourself. You can use this prompt for this purpose

Or you can go straight to the link and start generating.

You can also see examples of how to use prompts in my telegram group, where I share ready-made algorithms and image creation tweaks every day.

Download the results of your experiments and share them within the course. Remember that the main thing is your creativity and imagination! Good luck, and enjoy the process of creating unique images with artificial intelligence.