Loras

Little file, lots of power.

Loras are like extensions to the model, letting you get really specific things a lot easier. Like characters, poses, and art styles.

Installing a Lora

First step to using them is to head over to and sort you search to look for just Loras. (Other embeddings don't work. Textual Inversion, Hypernetwork, LoHA, etc. )

There are Stable Diffusion 1.X and Stable Diffusion 2.X Loras, just like models. You can only use SD1.X loras with SD1.X Models, and you can only use SD2.X loras with SD2.X Models.

Once you find one you like click on the big blue download button.

Be sure to keep note of any trigger words. Those will be needed to help activate the Lora's changes.

On a side note Loras with triggers like <loraNameV1:1> aren't needed, because that part is being handled by Artroom already, and is just howAutomatic1111 handles adding Loras.

After downloading the Lora you will need to move it into a sub-folder of the 'model weights' folder you have listed in Artroom

The folder has to be named "Loras" if it isn't already there.

Using a Lora

After putting it in that folder it should show up on the list of Loras in Artroom

Select it from the list and it will show up in the sidebar, allowing you to select what weight should be applied to the Lora.

Most Loras will tell you what weight to attempt using on their CivitAI page. Rarely should they be any higher than 1 weight. Usually a range of 0.3 - 0.7 will work best.

You can add as many as you want to the list.

But they might start conflicting and you will get bad results. Sometimes vaes will interfere with Loras too And extreme CFG values will also be worse when using Loras

Be sure to add the trigger from the download page, preferably at the start of your prompts, and it should now add that Lora to the generated image.

Also note that CLIP Skip can help when Loras aren't activating correctly. Clip Skip