About Teletype
Join
AI2GO SCHOOL
@ai2goschool
February 11

Lora AI-Generated Images — Simple Understanding With Examples

LORA AI models use machine learning to create stunning, unique, and sometimes surreal works of art that push the boundaries of traditional art forms. Lora itself represents the adjustment of your image.

So, to create an image, we need to initially determine the control point of the drawing (checkpoint), also called the drawing style or, more simply, the model. The model used can become the basis for the creation, since it already contains elements such as clothing, environment, characters, drawing style, etc. The model can be used without additional parameters and settings.

Here are some examples of models (checkpoints) that you can use in your work:

The presented models can be used without any add-ons and Lora.

But if for example you need:

  • Draw a specific character
  • Fix physique display
  • Add additional styles to your drawing
  • Create a unique character from several (yes, this is also possible)
  • Add complex elements, decorations or special clothing items

To solve all these problems and other similar problems, you may need Lora. Most often, in all works you can find at least one use of Lora in a query. But for example, in my works I always combine from 2 to 5 Loras, which in the end give me exactly the result I need.

Examples of using

1) Using extra Lora

As an example, we will look at some of my recent works, in which I tried to address the topic of using Lora. Let's take this work as a basis, in which I changed different Lora to obtain the required results.

https://www.seaart.ai/explore/detail/cn375jte878c73dcgo80?u_code=46RK7R

The following elements were used to create this work:

Prompt:

  • (masterpiece:1,2), max quality, cinematic highlight, wallpaper, 1girl, solo, violet evergarden, perfect eyes, blue eyes, blonde hair with ribbon, big breasts, Bruce Lee Style Yellow-Black Jacket, (jewelry, emeralds), (look at the viewer:1.3), simple background, white background

Model (checkpoint):

Lora:

You can repeat the work by following the link:

https://www.seaart.ai/explore/detail/cn375jte878c73dcgo80?u_code=46RK7R

As you can see, the image was created using Bruce Lee Style Yellow-Black Jacket but the image does not show Bruce Lee's clothing. How can this be done, you ask? And everything is quite simple. This Lora has a very low weight of 0.2 (importance level in the display). As a result, we get an image that will focus primarily on checkpoin and lore with a large weight. The correct selection of scales is very important and everything needs to be configured so that there are no conflicts.

In order to ultimately get the image we need, namely a girl in Bruce Lee's clothes, we need to make the following changes.

Prompt:

  • (masterpiece:1,2), max quality, cinematic highlight, wallpaper, 1girl, solo, perfect eyes, blue eyes, blonde hair with ribbon, big breasts, Bruce Lee Style Yellow-Black Jacket, (jewelry, emeralds), (look at the viewer:1.3), simple background, white background

Model (checkpoint):

Lora:

You can repeat the work by following the link:

https://www.seaart.ai/explore/detail/cn378d5e878c73dcltn0?u_code=46RK7R

As a result, we get this image

https://www.seaart.ai/explore/detail/cn378d5e878c73dcltn0?u_code=46RK7R

2) Proper use and combination of Lora

Now let's move on to practice and consider examples of combinations of different Loras in images. We will continue to work with the same image as before.

Example 1

https://www.seaart.ai/explore/detail/cn37a4te878c73dcp410?u_code=46RK7R

Everything that was used before was taken as a basis, but it was the Lora settings that were changed.

Prompt:

  • (masterpiece:1,2), max quality, cinematic highlight, wallpaper, 1girl, solo, perfect eyes, blue eyes, blonde hair with ribbon, big breasts, Orange Slingshot Swimsuit, (jewelry, emeralds), (look at the viewer:1.3), simple background, white background

Model (checkpoint):

Lora:

You can repeat the work by following the link:

https://www.seaart.ai/explore/detail/cn37a4te878c73dcp410?u_code=46RK7R

As we can see when adding a special Lora, which is responsible for the swimsuit Slingshot Swimsuit managed to get this result. All other Lora settings remained the same as before, but the result of the work is radically different. At the same time, do not forget about writing keywords in the prompt (in this case, description of clothing).

Example 2

https://www.seaart.ai/explore/detail/cn389cle878c73d20s00?u_code=46RK7R

Prompt:

  • (masterpiece:1,2), max quality, cinematic highlight, wallpaper, 1girl, solo, perfect eyes, blue eyes, blonde hair with ribbon, large breasts, (black bdsm latex), (latex breast bondage), (deep cutout breasts), (no sleeves), (jewelry, emeralds), (look at the viewer:1.3), simple background, white background

Model (checkpoint):

Lora:

You can repeat the work by following the link:

https://www.seaart.ai/explore/detail/cn389cle878c73d20s00?u_code=46RK7R

Again the settings don't change much, but when adding a special Lora the result is very different.

Example 3

https://www.seaart.ai/explore/detail/cn3tetle878c73efsg5g?u_code=46RK7R

Prompt:

  • (masterpiece:1,2), max quality, cinematic highlight, wallpaper, 1girl, solo, confused face, brush, perfect eyes, brown eyes, blonde hair with ribbon, gorgeous buff body, bunny ears, large buff slender breasts, (red Sexy Babydoll Fishnet Mini Dress), tightly fit red bunnySuit, fishnets stokings, (cutout breasts), (jewelry, emeralds), (look at the viewer:1.3), close mid shot, simple background, white background

Model (checkpoint):

Lora:

You can repeat the work by following the link:

https://www.seaart.ai/explore/detail/cn3tetle878c73efsg5g?u_code=46RK7R

In this example, I tried to combine two separate types of clothing

  • red Sexy Babydoll Fishnet Mini Dress
  • tightly fit red bunnySuit

And separately prescribed where to use it fishnets

  • fishnets stokings

All this is needed to manipulate the parameters of the connected Lora. So for example in this example I didn’t need to fishnets was used throughout the image. It was important for me that this parameter turned on exactly where I needed it. Two different clothing parameters gave a very beautiful and harmonious result. But this is only because I knew what could be combined and what would not cause conflicts.

To know what types of clothing there are, read my tutorial here

RU: https://teletype.in/@ai2goschool/clothes-and-accessories-in-ai-prompt

ENG: https://teletype.in/@ai2goschool/clothes-and-accessories-in-ai-prompt-en

But most often it is better not to combine clothes from Lora and Prompt, this can give unpredictable results.

Also as you can notice that Lora was removed The Sexy Muscular , I replaced it with prompt:

  • gorgeous buff body

You can not change and use everything the same as before, the result will be similar to mine.

Conclusion

In this tutorial we briefly looked at the capabilities of Lora:

  • what are Lora
  • how to use
  • how to combine
  • what problems could there be
  • how to solve emerging problems

If you want more tutorials like this or a better discussion of this topic, then write about it in the comments.

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.

AI2GO SCHOOL
February 11, 15:00
0 reactions
0 views