Skip to content

ManOrMonster/Hakkun-ComfyUI-nodes

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

35 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Installation

Just put hakkun_nodes.py into ComfyUI\custom_nodes

Also avaliable to install by Manager - https://github.com/ltdrdata/ComfyUI-Manager image

Drag and drop hakkun_nodes_workflow.png into ComfyUI to check use example for all nodes.

Example is using ComfyUI-Custom-Scripts

https://github.com/pythongosssss/ComfyUI-Custom-Scripts

Custom nodes

All nodes can be found in "Hakkun" category.

Prompt parser

image

Allows you to write whole positive and negative prompt in one text field.

You can write on multiple lines. All lines will be joined by "," and cleaned of unnecessary white space.

Special syntax examples:

Everything after @ will be used as negative prompt.

3d render @ drawing, painting

positive: 3d render

negative: drawing, painting

All lines starting with ! will be ignored.

All lines starting with ? will be randomly ignored with a default 50% chance.

You can modify the default 50% chance by specifying different percentages after ?.

?20% explosion

This will give you a 20% chance to include the word 'explosion'.

To randomly select from a list, use this format:

[dog|cat|horse]

You can nest these lists on multiple levels:

[town|city] full of [[fast|slow] [cars|[bicycles|motorbikes]]|[[loud|cute] [horses|[ducks|chickens]]]]

This will generate sentences like:

town full of cute horses
city full of fast bicycles

To adjust the chances of selecting each element separately, assign weights using the following format:

[*150*car|*30*boat|train|*80*bike]

If a weight is not specified for an element, the default weight of 100 will be used. Weights can be placed anywhere within the elements.

This also works with nested elements:

[[*10*pink|blue] bedroom*30*|*120*city at [day|*150*night] with [cars|trains|*10*rockets]]

In this example, a pink bedroom will be very rare.

There's also the option to insert external text in <extra1> or <extra2> placeholders.

Include <extra1> and/or <extra2> anywhere in the prompt, and the provided text will be inserted before parsing. If you don't specify these triggers but provide text, it will be pasted at the beginning by default (as positive).

All occurrences of em: will be replaced with embedding:.

If a line starts with ^, everything after that symbol and the next lines will be ignored.

All prompt between /* and */ will be ignored.

You can set the seed as an input to control randomness along with the rest of the workflow.

Using the debug output will provide you with all the information about the generated prompt.

Tags:

image

You can apply custom tags by tags string input or load text file from tags_file (full path)

Set tag by >>>YOUR_TAG + lines below as content (check workflow for example)

If you set tags by tags_file (full path) then it will be used instead of tags input.

You can use tags in tags only before you define those inner tags.

All tag lines will be joined into one line by , so:

  • don't use more that 1 negative @ by tag
  • don't use ! or ?

If you have pysssss nodes - setting tags file as default can be good idea (in comfyui settings):

image

image

Multi Text Merge

image

Allows to join multiple texts by specified delimiter. Put \n to use new line as delimiter. No need to keep any order. Empty inputs will be ignored.

Random Line

image

Will output random line

Random Line 4

image

Will output random lines from 4 textfields/inputs and join them by specified delimiter. Put \n to use new line as delimiter.

Calculate Upscale

image

Outputs target upscale float value based on input image height. Also calculates tile size (with and height) for tools like UltimateSDUpscale by specified number of horizontal tiles.

Image size to string

image

Outputs input image size in format: 512x768

Any Converter

image

Universal primitive type converter. If string cant be number it wil be 0

Image Resize To Height/Width

image

Resize image to desired width or height. Support batches. Uses lanczos.

Load text

image

Simplest load text file (somehow I could not find such stuff in other nodes that would work)

Load Random Image

image

Will return one random image ('.jpg', '.jpeg', '.png', '.webp') from passed directory and its file name (without extension).

About

For pull requests

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 100.0%