LCM: Latent Consistency Model

LCM (Latent Consistency Model) provides support for SD 1.5 and SD-XL models.

Installation

Download the LCM LoRA models and place them in your LoRA folder (models/lora or custom):

• For SD 1.5: lcm-lora-sdv1-5
• For SD-XL: lcm-lora-sdxl

Because both files have the same name, download them one at a time and rename each file before downloading the next.

Usage

1. Load your preferred SD 1.5 or SD-XL model that you want to use LCM with
2. Load the correct LCM LoRA (lcm-lora-sdv1-5 or lcm-lora-sdxl) into your prompt, for example: <lora:lcm-lora-sdv1-5:1>
3. Set your sampler to LCM
4. Set number of steps to a low number, e.g. 4-6 steps for SD 1.5, 2-8 steps for SD-XL
5. Set your CFG Scale to 1 or 2 (or somewhere between, play with it for best quality)
6. Optional: enable Hypertile and/or FreeU to improve output speed and quality
7. Generate!

Notes

• This also works with latent upscaling, as second pass/hires fix.
• LCM scheduler does not support steps higher than 50
• The cli/lcm-convert.py script can convert any SD 1.5 or SD-XL model to an LCM model by baking in the LoRA and uploading to Huggingface