You may not know how to use slider-LORA at all (Patreon)

In my daily creative work, I often use slider-LORA. For this reason, I have trained a large number of slider-LORAs and also published some, but I have found that few people use it correctly. The most common misconception is:

Why doesn't it have an effect on my image?

Because it is in the T2I model, that is, it is a LORA that exerts control when you generate images through prompts.

slider-LORA differs from ordinary LORA:

slider-LoRA can precisely control the changes of a certain element in the image by adjusting the weight. Taking the slider-LoRA for controlling the size of glasses as an example, several points can be observed:

The weight range of slider-LORA is quite large, usually between ±5, and sometimes it can go up to ±10.

The change in eye size varies linearly with the weight, so you can fine-tune the weight value precisely until you achieve your desired eye size.

Generally, the effects of positive and negative weights are opposite: a positive weight enlarges the eyes, while a negative weight reduces them.

Mentioning descriptions related to slider-LORA in the prompt will only create confusion:

For example, if the function of a slider-LORA is to achieve close-up shots, then the prompt should not include descriptions of camera lens models or high heels.

Multiple slider-LORA stacking is completely fine, so don't worry about the image collapsing. Making good use of this will take your creations to the next level.

Of course, you can also directly use the exclusive workflow I provided to reduce unnecessary troubles.

An excellent visual work has many problems to solve, while a LORA can only address one specific problem, and slider-LORA can address at most two. Therefore, you need to have the mindset of multi-LORA stacking. The following is a practical case of mine to help you understand how multi-slider-LORA works.

Depict a moment of a woman in a confined space:

For the narrow space, I chose the elevator. To convey the crampedness of the space, I used the slider-LORA that controls the number of people. After several rounds of image generation, I obtained a composition that I like.

It shows the narrowness and crowding of the space, but lacks aesthetic appeal, and some parts of the limbs have collapse issues. However, it's okay, as it's already sufficient as a good start.

To highlight the protagonist, I made her dressing a bit different, and I hoped she would be sexier. The most direct way was to accentuate her figure, so I overlaid a slider-LORA that controls size. (Note: I am overlaying LORA, not replacing it, and the same applies to subsequent operations)

The image finally became less cluttered, but I wanted the protagonist to be a young white woman, so I added the slider-LORA for controlling skin color and age

It's not bad, the image has started to change in the expected direction, but I'm still not satisfied. I hope she could look a bit thinner, so I've added the slider-LORA for weight control.

Great! This looks very close to the image I want. Now let's connect to upscale and see the results.

Fantastic! I hope you'll like it as much as I do! Maybe I can go shoot fashion spreads! (Friends in need of this please contact me! LOL)

See? The stacking of multiple slider-LORAs not only does not damage the image, but also gives you greater control power, thereby achieving predictable image results. This is the power of slider-LORA.

Thank you for reading this far. By now, you should understand how it works. I hope this brief sharing can inspire your creativity. If you find it useful, please give it a like, follow, and share it so that more people can see it.

All the aforementioned LoRAs are available to supporters. If you wish to become a supporter, it is recommended that you read this article first to gain a comprehensive understanding of the benefits entitled to supporters.