In jtkelm's wildcards script (repo link seems to be dead, but it should be similar to this), prompts are displayed both raw and unprocessed in the dialog box.

• Wildcard prompt: The raw text before it has been processed. This includes wildcards used
  • e.g. __season__
• Example: the processed prompt, with wildcards replaced with the selection for each item
  • e.g. winter
• Negative: unmodified negative prompt

While it might not be as helpful in "Generate forever" sessions without ensuring the wildcards are also exported, it could be helpful for single generations, and can help the user ensure their expressions are working as intended.

Though, after thinking on it, I wonder what we'd choose to display if this was implemented, since dynamic prompting allows for combinations and options.
Would we display the initial raw prompt without any evaluation?
What about recursive wildcard evaluation? Would that make this feature less useful?

I had a person wildcard with three names, but when I used it, I was getting images with prompts like "through a window," and "in space" and I realized it was taking from both the person wildcard and the personaction wildcard. I haven't noticed this until I updated to the .5 version of this script.

you used to use the script drop down but it is no longer found there. it also doesn't work either so it's not just globally enabled (plus even if it was, because it's not in the drop down it has no options or settings)

I have tested this in several ways.

It seems the current prompt limit when using wildcards is 125 (the value listed on the UI, next to the "Generate" button); after this value, the wildcards stop working.

Or at least, this is happening on my end.

Describe the bug
A clear and concise description of what the bug is.

normalcy ex) {3$$__paser__$$__values__}
-> __values__ __paser__ __values__ __paser__ __values__
-> [ a and b , c ],[ b , c and c ],[ a and b and c ], ...

bug ex) {3$${and|,}$$__values__}
-> {3$$and$$a}
-> [ a and a and a] , [ b and b and ]

I think it needs to be upgraded

Version used
What version of the extension are you using? You can see it next to Dynamic Prompt v[VERSION] in the scripts dropdown.

under 0.25.0

To Reproduce
Steps to reproduce the behaviour, include the prompt you used if applicable:

1. Go to '...'
2. Click on '....'
3. Scroll down to '....'
4. See error

Also mention which checkboxes have been set.

Expected behaviour
A clear and concise description of what you expected to happen.

Additional context
Add any other context about the problem here.

White mode

Dark mode

Love the script! Question - is it possible to adapt Dynamic Prompting into an extension?

This is absolutely beyond amazing, thank you so much for its development,
however updating to try and take advantage of the new features has broken my previous usage of the script.

With the wildcards locked to seeds? pressing 'generate' with wildcards now always returns the same result. This is a problem because I took advantage of wildcards primarily because they meant I could find a good seed with the surrounding prompt, but vary it.
For example I could press generate and randomize the characters features with wildcards but hold the rest of their construction.

If this was meant as a 'fix' for it not being locked to the seed before, could this please be returned with a toggle, e.g. 'separate wildcard randomization from seed'.

I tried to install by terminal and also via extensions in Gradio by it's not taking any __textfile__ . I do have them in the wildcards folder downloaded by git.
Ubuntu 20

Combinatorial mode does not work with ranges.

Example: {1-2$$red|blue}
Expected:

• red
• blue
• red,blue

Actual

• 1-2$$red
• blue

Hello!

While I do have it installed (used the git clone command into my extension directory) and it shows up in the extensions tab :

It does not show anywhere else in my UI. No tab no extra option in the settings, nothing.

I am using a fresh colab so it is not due to residual files.

Best regards,

Really loving this script- Is there any way we can have a wildcard file with seeds in it? In IMG2IMG I tried with prompt "(blah blah), Seed: seedlist" and it used the seed listed in the "seed" field in the UI, instead. I would really love a way to control the seed used via prompt and, of course, use a wildcard file for them....if any of that's at all possible. Thanks!

{{2-3$$__artist__}}

Just tried to use the above and it failed. Script seems to demand I do:

{{2-3$$__artist__|__artist__|__artist__}}

I'd also like to do things like:

{{2$$__artist__|__movement__}}

to give a mix of two artists, two movements, or one of each; without having to put each twice.

I can't tell if I've done the setup wrong, or if I'm running into a bug. Installed from the Extension's page, and it's saying I'm up-to-date.

I've got my files in the default WILDCARD_DIR, like such:
scripts\wildcards\clothing\outfits.txt
scripts\wildcards\clothing\shoes.txt
scripts\wildcards\clothing\tops.txt
scripts\wildcards\clothing\bottoms.txt
scripts\wildcards\test.txt

test.txt contents (all on new lines, edited with Notepad so it probably has the Windows formatting)

Summer
Autumn
Winter
Spring

When I type in the prompt a lake in __test__ I get the auto complete suggestions showing up as soon as I type __, which is a good sign everything is in the right directory. If I select one of the files, it lists out all the options in it so I can manually pick one, which again makes me think I'm set up correctly.

However, when I go to run it, I get this error in the command prompt:
File E:\Documents\AI\ImgGen\stable-diffusion-webui\extensions\stable-diffusion-webui-wildcards\wildcards\test.txt not found for the __test__ wildcard.

It gets even worse when I try to use the clothing ones:
File E:\Documents\AI\ImgGen\stable-diffusion-webui\extensions\stable-diffusion-webui-wildcards\wildcards\clothing/outfits.txt not found for the __clothing/outfits__ wildcard.
Notice the full path there. It's swapping between \ and / for the directory. Not good!

The lake never generates with anything other than a vibrant green trees and grass around it (take your pick on Summer or Spring), and the text under the image generation is always this (minus the seed changing)

a lake in test
Steps: 28, Sampler: Euler a, CFG scale: 11, Seed: 945332860, Size: 512x768, Clip skip: 2, Wildcard prompt: a lake in __test__

I assume if the wildcards were working one of the prompts would replace test with one of the seasons?

I am getting this error message while using Magic prompit :

AttributeError: 'NoneType' object has no attribute 'keys'

what i want to achieve for example if my prompit is like this { cat | dog | chicken } in the garden and are { happy | sad }

what i want to achieve is i get prompits like this

cat in the garden and are happy
cat in the garden and are sad
dog in the garden and are happy
dog in the garden and are sad
chicken in the garden and are happy
chicken in the garden and are sad

and the text file will be the same as the mentioned prompits and when it finish it stops , how can i achieve that ?

Nice extension but i think it doesnt work as documented

I installed it (as extension)
chose Dynamic prompting v0.13.7 (in Script section)
checked box Combinatorial generation
set Combinatorial batches to 4
prompt: test {dog|cat|rabbit}

according to docs i should have 3 images with combination of 3 different animals
but i get 4 images with one of animals every time (same animal for all 4 images)

am i doing something wrong?

don,t use [] plz, like this:[2$$opt1|opt2|opt3], it will conflict with webui Attention/emphasis

I don't understand why the script is not working for me.

Error loading script: dynamic_prompting.py Traceback (most recent call last): File "~~~/stable-diffusion-webui/modules/scripts.py", line 159, in load_scripts exec(compiled, module.__dict__) File "~~~/stable-diffusion-webui/scripts/dynamic_prompting.py", line 52, in <module> class WildcardManager: File "~~~/stable-diffusion-webui/scripts/dynamic_prompting.py", line 65, in WildcardManager def get_files(self, relative:bool=False) -> list[Path]: TypeError: 'type' object is not subscriptable

I'm loving this script! So much fun to mess around with!

Is it possible to send the generated prompt in txt2img over to inpainting / img2img? Currently I'm just copying and pasting the generated prompt into the in-paint tab. It's only a minor annoyance is all. Having a button to send the prompt and settings would be cool!

Thanks again!

With a large number of wildcards, it can be unwieldy to maintain several large files in a single directory.
It would be nice if this script could scan recursively for wildcards in file structures of arbitrary depth.

Here is how I am handling it in my repository, using rglob.

Also note that other repositories such as a1111-sd-webui-tagcomplete have already implemented support for this feature; a similar update to dynamic prompting would complement this nicely.

To help improve readability, I also changed the script text element to display paths relative to WILDCARD_DIR, and made the list scrollable, to prevent a large number of wildcard files from expanding the page uncontrollably.

Happy to make a PR if the use case is valid and the changes (or something similar) sound acceptable.

For example (muscles:1.1) and (muscles:1.9) will give you wildly different results.

with text files I've tried (muscles:1.__num__) with the num text file being number 1 through 9 so output being (muscles:1.#).

I've also tried (muscles1.1) through (muscles1.9) in a textfile then using __muscles__ and that didn't work. I also tried without parenthesis and it didn't work. Bascially every possible way to do it in combination with pulling from a text file did not work.

I looked through the code and can't understand why it's not working when ( ) + prompt weights are involved.

Loving this script. Is there a way to reiterate wildcards one by one?

version 7cfd16e

expected: using a __filename__ should randomise per batch

actual: it does not

similarly {prompt1|prompt2|prompt3} will not randomise per batch

I saw in the notes that i can use
If a wildcard is provided, then a new prompt will be produced for every value in the wildcard file. For example: My favourite season is seasons

will produce:

My favourite season is Summer
My favourite season is August
My favourite season is Winter
My favourite season is Sprint

But this doesn't work. I tried using combination number, batch number, as well as * but it justs picks the things randomly from the file.

I think it should work like below

file contains 3 colors
red
black
white

my favorite color is MyColor (Some modifier) and it should give me 3 images with

my favorite color is red
my favorite color is black
my favorite color is white

Increasing batch size does nothing with this script on. Tried using no wildcards, reducing size of image, different samplers, etc. The only thing that makes batch size work again is turning off the script. Batch size increases overall speed of generating many images in cards with lots of vram so it's very useful.

I have a 3090 using voldy commit hash f49c08ea566385db339c6628f65c3a121033f67c

In the readme examples for Combinatorial Generation, it demonstrates multiple {} contents, and a single wildcard, however it mentions nesting in both directions.
But as soon as I have more than 1 wildcard, it silently breaks for me and I have to restart.

If that's an intended limitation because of the excessive amount of possibilities with multiple wildcards, though that confuses me with "This upper limit ensures that you don't accidentally create a template that unexpectedly produces thousands of images" being mentioned. I wonder if you could potentially target a specific wildcard list for the combinational generation.
For example {wildcard} would have the generation go through that, with randomized elements of other wildcards. That way you could have a wildcard you desire seeing all the elements of, and set your batch to its size, and give another form of utility for this extension.

I have a file of random "careers" for characters. I'm hoping to reuse the random choice to indicated both the career and the uniform the character is wearing but I am not having much success.

I've tried:

A __career__ wearing a __career__ outfit but what I end up with is something like:

A teacher wearing a fishmonger's outfit [NOTE: not that there's anything wrong with that in real life!] when I wanted a teacher in a teacher's outfit

I thought I'd be clever and see if using "$1" or "\1" as a regex group holder would work, but that failed miserably :)

You tried to be exhaustive with the artists, I understand very well your aim but you fell into the same trap I fall myself some weeks ago when I tried to create SD artits database. The fact is only a very small subset of all artists existing has been used the train Stable Diffusion. The list of artists should be reduced to the one SD has been trained with, otherwise you kinda just produce veru unprecise artist style.

I advice you to read this article
https://waxy.org/2022/08/exploring-12-million-of-the-images-used-to-train-stable-diffusions-image-generator/

Let me take an example I took your french_mannerist_painters list
Ambroise Dubois, Antoine Caron, Francesco Primaticcio, François Quesnel, Luca Penni, Marc Duval, Martin Fréminet, Master of Flora,
Toussaint Dubreuil

I render six images with each of them using that prompt
a deer in the forest by
a close-up portrait of a woman by
an old king by
merchants in a city by
a bird on a tree branch by
the last judgement by

Here are the kind of result i got

Duval Marc

Dubois, Ambroise

Penni, Luca

Quesnel, François

I can tell you SD doesn't know any of those artists style, maybe he knows they are renaissance artists not even but that's all

Now I render the same prompt with renaissance artist I know SD is aware of

Bellini, Giovanni

** Masaccio**

Parmigianino

van Eyck, Jan

Veronese, Paolo

For your artist lists to be really useful you will need to change your mind and reduce your lists dramatically. From my personal research, so far I have not found more than 50 renaissance artists known from SD and still for certain I have some doubts

Now the question is how .. i'm not aware of any "certified artists" list of SD. Probably the first step is to reduce to the artist list that others found out like the 1800 artists described into the article mentionned above.

Hope this won't kill your day.

I'm getting these errors when loading StableDiffusion, not sure what they're affecting specifically, so i thought I'd share them with you to see if you know more about this. Thanks in advance for your assistance in this.

So, I get it, you don't want the user to accidentally generate thousands of prompts and all, but, really, that's on the user.

The way you have it set now, I can have, at max, 32 combinations, unless I am missing something here... I could be, but, due to my processing limitations, I can't have the batch size very high.

So if I am doing everything right, I can only do a batch size of 2 and the max for batch count (Unless there is someway to change that) is 16. Now my problem is I have 20 objects and 30 situations, objects in one file, situations in the other, and I want to combine them and that's like 600 combinations.

Can we please get a better way to limit how many we can generate at a time.
Or, if there is a way I can do it with the current system, can we get an explanation as to how.

Thanks -J

Would it be possible to make it so that dynamic prompts are read correctly from inside of the wildcard .txts? For example, I have a wildcard file I use to pick random outfits and would like to attach specific randomly chosen backgrounds to each outfit inside the .txt, but right now the images generate a combination of all the tags inside of { | | | } markers rather than selecting one.

It would be nice if the extension could be globally enabled using a toggle in the settings instead of having to choose from plugins. Less cumbersome and also allows you to use dynamic prompts within other scripts.

Thank You for updating the script into an extension - great improvements and possibilities!

Noticed (in one of the closed issues) that it might be tricky to make it work with other scripts / integrate it into SD. I'm interested in X/Y plot in particular (to allow option tweaking, like changing hypernets, steps, CFG, etc.).

Is it possible with the current iteration of Dynamic Prompting / possible at all?

Describe the bug

INFO:dynamic_prompting.py:Prompt matrix will create 8 images in a total of 8 batches.
ERROR:dynamic_prompting.py:Failed to write prompts to file: 'charmap' codec can't encode character '\u0142' in position 1208: character maps to

Version used
Dynamic-prompt extension 0.25.0

think is due to my artist.txt wildcard file that contain nord-european painter with Ä (142 in my charmap) in their name.
some google search suggest it correspond to this polish char: ł

I'd like to try something like this

A {male|female} {elf|orc|goblin} {warrior|priest|wizard} {walking in the {forest|desert}|running in the {jungle|arctic}}

but the script does not give me the alternatives at the end. I end up with results like:

A female elf priest {walking in the forest|running in the arctic} ... so it's picking the variables but not recursively picking the setting. Is that possible to do?

Thanks,

TB

If you write a prompt like:

a painting of [Eva Green|Scarlett Johanson|deborah ann woll|Emily Blunt]

and activate Magic Prompt, the resulting rendered prompt will be:

a painting of [Eva Green, Scarlett Johanson, deborah ann woll, Emily Blunt]

switching the "|" with a comma ",".

That will result in multiple subjects instead of an alternated one.

Version: should be a729de9 (it's not in the scripts dropdown)

To Reproduce
see above.

Only magic prompt checkbox has been set.

Cheers.

it would be nice if we had a way to add a weight to lines of a wild card instead of copying and pasting that same line multiple times
example:

hot[9]
cold[9]
freezing[3]
burning[3]
on fire
frozen

this could make hot 3 times as likely as freezing and freezing 3 times as likely as frozen
in addition to this, something we can put on the line that is nothing, for the chance to get no selection and control that chance with a weight

Currently wildcards are not possible given a limitation in stable-diffusion-webui. An issue was created here: AUTOMATIC1111/stable-diffusion-webui#2814

originally requested by Adil3tr

Hi,

The extension is a masterpiece and I love it.

I am trying to add the "wildcard_dir":"D:\\wildcards" in stable-diffustion-webui/config.json

but the error shows as below:

Error loading script: dynamic_prompting.py
Traceback (most recent call last):
  File "C:\stable-diffusion-webui\extensions\sd-dynamic-prompts\prompts\, line 16, in ensure_directory
    self._path.mkdir(parents=True, exist_ok=True)
AttributeError: 'str' object has no attribute 'mkdir'

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "C:\stable-diffusion-webui\modules\scripts.py", line 170, in load
    exec(compiled, module.__dict__)
  File "C:\stable-diffusion-webui\extensions\sd-dynamic-prompts\scripts\y", line 133, in <module>
    wildcard_manager.ensure_directory()
  File "C:\stable-diffusion-webui\extensions\sd-dynamic-prompts\prompts\, line 18, in ensure_directory
    logger.exception(f"Failed to create directory {self._path}")
NameError: name 'logger' is not defined

seems is python missing a logger library.

logger.exception(f"Failed to create directory {self._path}")
======
print (f"Failed to create directory {self._path}")

I try to fix it by replacing logger with print, it works for a while, but error shows in other lines

in pick_wildcards
    return re_wildcard.sub(lambda x: self._replace_wildcard(x), template)
  File "C:\stable-diffusion-webui\extensions\sd-dynamic-prompts\prompts\generators\randomprompt.py", line 100, in <lambda>
    return re_wildcard.sub(lambda x: self._replace_wildcard(x), template)
  File "C:\stable-diffusion-webui\extensions\sd-dynamic-prompts\prompts\generators\randomprompt.py", line 78, in _replace_wildcard
    wildcard_files = self._wildcard_manager.match_files(wildcard)
  File "C:\stable-diffusion-webui\extensions\sd-dynamic-prompts\prompts\wildcardmanager.py", line 33, in match_files
    WildcardFile(path) for path in self._path.rglob(f"{wildcard}.{constants.WILDCARD_SUFFIX}")
AttributeError: 'str' object has no attribute 'rglob'

If I use the default WILDCARD_DIR, everything works fine.
May I know can the extension support change wildcard_dir to D:\ drive?

i would like to pull full list included in folder not specify what file exactly

I am not saying anything, but it is broken:

Traceback (most recent call last):
  File "D:\Emergency\Stable Diffusion WebUI\modules\ui.py", line 185, in f
    res = list(func(*args, **kwargs))
  File "D:\Emergency\Stable Diffusion WebUI\webui.py", line 54, in f
    res = func(*args, **kwargs)
  File "D:\Emergency\Stable Diffusion WebUI\modules\txt2img.py", line 45, in txt2img
    processed = modules.scripts.scripts_txt2img.run(p, *args)
  File "D:\Emergency\Stable Diffusion WebUI\modules\scripts.py", line 296, in run
    processed = script.run(p, *script_args)
  File "D:\Emergency\Stable Diffusion WebUI\extensions\sd-dynamic-prompts\scripts\dynamic_prompting.py", line 137, in run
    prompt_generator = MagicPromptGenerator(
TypeError: MagicPromptGenerator.__init__() takes 2 positional arguments but 4 were given

Steps to reproduce:

1. Turn MagicPrompt on
2. Write some initial text (not sure it is needed)
3. Start generation

Why do you send it more arguments than needed?

Here is my prompt :

on stage, club, indoor, table, seats, chair, bar, standing microphone, realistic lighting, colored lighting,
highly detailed, digital painting, concept art, smooth, sharp focus, sharp focus, illustration, hyperdetailed, hyperreal, trending on artstation
ilya kuvshinov and range murata,

Here is what is send to the generator :

on stage, club, indoor, table, seats, chair, bar, standing microphone, realistic lighting, colored lighting,
highly detailed, digital painting, concept art, smooth, sharp focus, sharp focus, illustration, hyperdetailed, hyperreal, trending on artstation
ilya kuvshinov and range murata, holding

The only thing added is "holding"... I had that multiple times, seems like the longer the prompt is, the less magic prompt works. Maybe because of some arbitrary limitation, but this is annoying really fast.

Here is the PNG if you want to extract the data from it. https://i.imgur.com/cJVCyAG.png

I wish we can enable the plugin to have no regards about the prompt lenght or something, maybe the plugin can have some sort of button you clic and it give you random parameters that you can choose to add to your prompt ?

Describe the bug
When I updated the extension, I got this error in the terminal.

ImportError: cannot import name 'FeelingLuckyGenerator' from 'prompts.generators' (F:\repos\auto111\stable-diffusion-webui\extensions\sd-dynamic-prompts\prompts\generators\__init__.py)

Version used
Dynamic Prompts doesn't show in the scripts dropdown. But I just updated.

To Reproduce
Steps to reproduce the behaviour, include the prompt you used if applicable:

1. Updated extension
2. Saw error in terminal

Could be helpfull something like the grid generation but using the prompt, something like:

%CFG_SCALE=29|30|31|32% %SAMPLIG_METHOD=Euler a|Euler% etc etc.

And maybe better, something like ranges as the number of parameters to print:

%CFG_SCALE=$$29-35% //This could be also applyed to the basic prompting, but is not helpier than the usage with generation params

The exstion is installed but is no longer showing in the drop down menu after 0.15< just updated from 0.14 which did show in the menu

got this error when running, doesn't seem to work.

Many thanks to see there are many updates recently.
I see the script are changed into an extension since 0.11.
But the UI is still in the "script" drop-down menu. Maybe it should move to the setting page？

I want to use it with an other script to run repeatly, but currently the Combinatorial mode is not working with batchs.

I do apologise if this is the wrong place as I'm not sure what implements this, but when writing a prompt, I can use the tab key to autocomplete wildcards. This is very useful, except it will auto-complete whichever one is the top one, rather than filling in only what is "common" to the current possibilities. This means I have to type out a considerable amount before I can auto-complete usefully.

For example, if the three files are "foofoo", "foobar" and "foobaz". I would expect the following:

1. f<TAB> becomes "foo"
2. foob<TAB> becomes "fooba"

Currently, f<TAB> would just autocomplete whichever happens to be first in the list of possibilities.

I love wildcards and almost always use them, but there's two functions that can't seem to use them.

The first is batch processing. The wildcards script works but only uses one value for all images in the whole batch count. This dynamic prompting script actually just gives me an error when I try to do batch processing. (batch processing is in the img2img tab, it lets you run img2img on an entire folder of images with the same prompt and values). I would love to be able to use batch processing with dynamic prompting, and a unique combination of wildcard values in every individual image generation in the batch.

The second is wildcard/dynamic negative prompts. I sometimes use wildcards as a way of just adding variability, and I would love to be able to have "(adjective)" as my negative prompt to generate more kinds of variation in my generations. Right now it doesn't seem to recognize a wildcard in the negative text book, and the generated image's txt file just says "adjective" instead of assigning it a value.