santa hat
deerdeer nosedeer glow
Sign In

Stable Diffusion - Parameters to speed up the image and video generation

Stable Diffusion - Parameters to speed up the image and video generation

Objective

This article is to share some information about the Stable Diffusion A1111 parameters that I setup comparing with the default configuration.

Where to find all the parameters:

https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Command-Line-Arguments-and-Settings

How to set the parameters

Edit the file : go to <Stablediffusion folder>\stable-diffusion-webui\webui-user.bat

This is my current configuration :

@echo off

set COMMANDLINE_ARGS= --xformers --upcast-sampling --opt-split-attention --disable-safe-unpickle --api
call webui.bat

I have Nvidia 3060 12GB graphic card.

With a default configuration to generate the image below with same settings it takes 13.3sec.

With my current configuration it takes 8.3 sec

The quality is the same.

LCM

I notice that If I use the LCM with lora and few steps with my current configuration it takes longer

With the default settings it takes 3.8sec

With my settings it takes 2.7sec.

AnimateDiff

To generate a video with 8 frames.

With the default settings it takes 4 min. 28.4 sec.

With my settings Time taken: 54.0 sec. => This is great ....

Find attached to the article the image and the video generated.

Write in the comments your configuration and let me know if you can improve your settings in a better way.

Thanks for reading.

18

Comments