Lora

Lora is an abbreviation for Low-Rank Adaptation. It enables users to swiftly fine-tune diffusion models using low-rank adaptation technology

LoRAs (Low-Rank Adaptations) are smaller files (anywhere from 1MB ~ 200MB) that you combine with an existing Stable Diffusion checkpoint models to introduce new concepts to your models, so that your model can generate these concepts.

These new concepts generally fall under 1 of 2 categories: subjects or styles.

Subjects can be anything from fictional characters to real-life people, facial expressions, poses, props, objects & environments.

Styles include visual aesthetics, art styles, and artist styles.

Note that these LoRAs only compatible with SD1 and SDXL models only.

Available SD 1.X LoRAs

• 0mib3_v10

• 3DMM_V12

• age_slider_v20

• arcane_offset

• AstralMecha

• epi_noiseoffset2

• eye_size_slider_v1

• GrayClay_V1.5.5

• hair_length_slider_v1

• hipoly_3dcg_v20

• linevichit3-v10

• more_details_v10

• muscle_slider_v1

• room2

• StealthMecha

• weight_slider_v2

• zoom_slider_v1

Available SDXL LoRAs

• 3lectronics_v10

• blu3print_v10

• c0nst3llation_v10

• cyborg_style_xl-v10

• LogoRedmond_v10

• ParchartXL-v12

• pixel-art-xl-v1.1

• StickersRedmond

• t-shirt_design_v1.0

• v0ng44g_p0rtr14t_v10

• v0ng44g_p14nt1ng_v10

• w00len_2500_v10

• xl_more_art-full_v1

How To Use Them?

Simple! Just add the LoRA along with your prompt. Here is a little breakdown that describes how LoRAs work and how to use them properly.

<lora:LORA-FILENAME:WEIGHT>

• LORA-FILENAME is the filename of the LoRA model mentioned above this article. Copy paste the file name of the LoRA that you want and continue

• WEIGHT is how strong you want the LoRA to be. The range is 0-1; 0 is the same as deactivating the LoRA, 1 is maximum strength. With many LoRAs, a weight of 1 can be overpowering to the point of creating poor results, so experiment with lower values like 0.5 to 0.8

Here is an Example: <lora:more_details_v10:0.2>

(Detail Tweaker is a little special. It goes in 2 directions, and accepts anything between -1 and 1. -1 will remove details, and 1 will add details.)