-
Notifications
You must be signed in to change notification settings - Fork 1
Quick Features Explanation
Most basic: Just use the defaults in settings
node .
make sure you copy default-user-settings.js over to user-settings.js because this is where you will add your user settings.
Use user-settings.js to override settings in these files:
settings.js
image-settings.js
upscale-settings.js
Command line arguments always have the final say
You can also have presets which can change any settings by adding a preset in the ./presets folder
and running
node . --presets "preset1,preset2,preset3"
Presets can be useful for having different sizes, different prompts, or different negative keywords
These are loaded after settings but before command line arguments
Expansions in ./expansions take a string of text and dump it into the prompt when referenced.
This is good for commonly used prompts wording.
You reference expansion files with <filename> in the prompt
<rays> => god ray, light shaft, volumetric lighting
<flower-pic> => {flower}, {flower}, {artist} => Calla Lily, Daisy, Hans Hofmann
Everything in ./lists is a list, a list is a long text file with elements on new lines.
One is randomly chosen when referenced.
You reference lists with {filename}
{flower} => Grape Hyacinth
By default, lists are guaranteed to be unique entries in a single prompt and non-repeating
Dynamic code in ./dynamic-prompts that is executed on each prompt, these form a chain, the prompt is
passed to the first dynamic prompt, then the output of that dynamic prompt is
passed to the next and so on until all dynamic prompt modules have been used.
Useful for dynamically generating certain prompts or modifying existing prompts such as adding new features.
landscape => Random landscape prompt
Within the prompt, you have access to these special lists which aren't real list files.
{keyword} Picks a keyword from whatever the currently selected dictionary is
{artist} Picks an artist from whatever the current selected artist dictionary is
{random} Picks a keyword from any non-artist list file in the data directory
{random-artist} Picks a keyword from any artist list file in the data directory
{salt} Replace with a random number in brackets
--prompt <prompt> The prompt to use
--negative-prompt <prompt> The negative prompt to use
--hide-prompt <true/false> Don't print the prompt to the console
Adding {salt} will add a random 10-digit number to your prompt in square brackets
to add variation to the prompt. Once this number is added to the prompt, it will
automatically be re-updated everytime the prompt is re-used.
If you want to append a salt to the prompt automatically without using {salt}
then the argument --prompt-salt will do it. It will skip over images that
already have a salt to prevent duplicate salting however if you wish you can add
multiple salts like so {salt} {salt} etc... and all of them will get updated
everytime.
You can optionally have the salt increment rather than be random, if so, use the
--prompt-salt-start <number> command line argument.
--prompt-salt <true/false> Auto-add prompt salt to the end of every prompt if not already there
--prompt-salt-start <number> Start the prompt salt at a certain number and gradually increment it with each prompt
This is mainly used only when {prompt}, {prompt-random}, or {prompt-danbooru} is used in the prompt.
{prompt} will randomly select keywords from the current dictionary and artist dictionary
{prompt-random} will use words from all dictionaries
{prompt-danbooru} will use words from the danbooru dictionaries but this may not be as random as just selecting the danbooru dictionaries and using the normal {prompt}
In your own prompt, {keyword} and {artist} will use the current dictionary
while {random} will use all dictionaries
--count <number> Number of keywords to generate randomly
--max-count <number> Randomize the amount of keywords, and this is the max amount
--prompts <number> Number of prompts to generate
--emphasis <true/false> Randomly emphasize / de-emphasize prompt keywords
--emphasis-chance <0.0-1.0> Chance it will aquire an additional level of emphasis
--emphasis-max-levels <number> Max level of emphasis
--de-emphasis-chance <0.0-1.0> Chance it will be de-emphasized rather than emphasized
--use-artists <true/false> Whether to allow artists to be used at all
--min-artists <number> Minimum number of artists to add
--man-artists <number> Maximum number of artists to add
--artists <name> Artist dictionary/list filename {artist} will reference this list
--dict <name> Keywords dictionary/list filename {keyword} will reference this list
--generate-images <true/false> Ask WebUI to make images
--upscale-images <true/false> Ask WebUI to auto-upscale the new image it makes
--list-entries-once <true/false> Allow list entries to be guaranteed unique
--reload-lists <true/false> Don't persist list entry uniqueness across all prompts in a batch
--list-files <path> Directory where lists files are
--expansion-files <path> Path to the folder for expansion files
--preset-files <path> Path to the folder for preset files
--dynamic-prompt-files <path> Path to the folder for dynamic prompt files
--image-save-to <path> Path to the folder for generated image files
--dyn-prompts <files> Add in new dynamic prompt modules to the start of the chain
--all-dyn-prompts <files> Customize the entire dynamic prompt module chain, you will need to specify existing ones
--file-variations Load generation settings of a certain file to make variations, command line arguments can override settings
--upscale-file Upscale a particular image, you can use upscale arguments to customize it
--webui-url <url>
Self-explanatory
--image-sampler <sampler>
--image-steps <steps>
--image-width <width>
--image-height <height>
--image-restore-faces <true/false>
--image-denoising <0.0-1.0>
--images-per-prompt <number>
--image-cfg <number>
--image-seed <number>
--image-subseed-strength <0.0-1.0>
Remember, if your wanting to upscale an existing image use --upscale-file,
otherwise it will auto-upscale a newly generated image.
--upscale-save-before <true/false> Save image before auto-upscaling
--upscale-to-size <true/false> If true, will upscale to exact dimensions, otherwise will upscale by a multiplier
--upscale-by <number> If --upscale-to-size is false, this is the multiplier number
--upscale-to-width <number> If --upscale-to-size is true, this is the upscale width
--upscale-to-height <number> If --upscale-to-size is true, this is the upscale height
--upscale-crop <true/false> If --upscale-to-size is true, auto-crop to upscale dimensions
--upscaler-1 <name> Name of 1st upscaler
--upscaler-2 <name> Name of 2nd upscaler
--upscaler-2-weight <0.0-1.0> Weight of the two upscalers
--upscale-gfpgan <0.0-1.0> GFPGAN percentage to use on image
--upscale-code-former <0.0-1.0> Codeformer percentage to use on image
--upscale-code-former-weight <0.0-1.0> Codeformer weight compared to GFPGAN
--upscale-late-face-restore <true/false> Fix faces after upscale (suggested to not use this)