Several people have asked for a copy of the wildcards that I use in my prompts for images and videos. Here they are!
Please consider these wildcards as just a starting point for your own prompt construction. Edit, delete, add, and make the wildcards your own.
Installation:
Unzip the wildcards .zip file and you'll have a "darkroast" folder. Place that folder inside the "wildcards" directory of the random prompts utility of your choice.
AdaptivePrompts in ComfyUI:
ComfyUI\custom_nodes\comfyui-adaptiveprompts\wildcardsDynamicPrompts in ComfyUI:
ComfyUI\custom_nodes\comfyui-dynamicprompts\wildcardsDynamicPrompts in Automatic1111:
stable-diffusion-webui\extensions\sd-dynamic-prompts\wildcardsDynamicPrompts in Forge:
stable-diffusion-webui-reForge\extensions\sd-dynamic-prompts\wildcards
Most of the image and video AI systems out there are based on one of those two tools, so the directories should be similar the tool you're using. See notes below for info on AdaptivePrompts and DynamicPrompts.
Here are a few points that might help you use or change the wildcards...
Each wildcard file has a brief description at the top to let you know what's included, so I'm not going to repeat that all here.
The prompts are formatted to work with AdaptivePrompts. Most of the prompt files will also work with DynamicPrompts. For the simple syntax included in these wildcard files, the only difference between those two prompt utilities is in how they handle the file glob syntax for including multiple files in a directory. If you want to use a catch-all prompt like "all_activities", for example, you'll need to edit that file before you can use it in DynamicPrompts. Other than that syntax difference, these prompts will work equally well in both.
I prefer to work with AdaptivePrompts because it knows how to reload the wildcard files without a comfyui restart. If you edit a wildcard file and save it, AdaptivePrompts will pick up your change on the next render, which is a big time saver when you're trying things out.
I've found that AdaptivePrompts doesn't seem to like an empty prompt option and will sometimes do strange things, so I typically use a _ character in place of an empty prompt. This is helpful for a section of text that should appear sometimes, but doesn't have alternate text. Like for example:
the man is driving a car {_|with a sunroof}.
In this prompt, the sunroof is optional, but there's no alternate prompt. It's either the sunroof or nothing. The _ is used as the "empty prompt" option because it doesn't seem to affect Wan's text parser at all, so it's safe to include in your prompt without changing the output.My prompts are geared towards female characters, so you'll have some editing to do if you work with men in your videos. Sorry.
All of the prompts are written without a period at the end so they are safe to use in a sentence with other text.
Examples:
{wide|medim|close-up} shot. natural light, warm light, rim light.
a woman wearing a t-shirt and jeans is __darkroast/all_activities__. she smiles at the camera.
__darkroast/cameraangle-wan__.
the camera pushes in for a close-up of her upper body.{wide|medim|close-up} shot. natural light, warm light, rim light.
a woman wearing a __darkroast/simple_colors__ {bikini|one-piece swimsuit} is {__darkroast/activities/beach_pool__|__darkroast/activities/home__}.
__darkroast/cameraangle-wan-nobehind__.
the camera pushes in for a close-up of her face.Description
FAQ
Comments (10)
Stupid question but where do I put the wildcard files?
.....extensions\sd-dynamic-prompts
I'll add that info to the wildcard model description, so have a look for it there.
@darkroast175696 much appreciated!
I was wondering about wildcards for a long time. So there are nodes that give you a drop-down to choose from or how does that work?
No, that's not it. There are nodes that will process the text of your prompt provided the prompt uses a certain syntax to list multiple options.
There are two ways to do it: random and let's call it sequential.
Random: several choices are given in the prompt and the node just picks one at random so you never know which one you'll get when you render the image or video.
Sequential: several choices are given and if you let the system keep rendering, it will go through all the options in order, then repeat until you tell it to stop.
Those are both useful in different situations. For sequential, I like to set up maybe a half dozen scenes that I know will probably take several tries before I get a result I like. I put them all into the prompt with the right syntax and just let it run overnight while I'm sleeping.
The random ones are nice to give you a surprising combination that you might not have tried. It's also really helpful when you're making a LoRA and you want unusual combinations of things in your prompt to see if your model will work or break.
So let's say you have a prompt like:
the boy {kicks|throws} the {black|white} ball.
If you let that run at random, you will get "the boy kicks the black ball" sometimes and other times a different combination. Now let's say you have 20 different colors because you're making ads for a sporting good company or something. You don't want to have all that text in your prompt, it's a huge pain. So instead you can take all those colors and just put them in a text file. The prompt will pick a random line from the text file instead of choosing a random option from the { } parts of the prompt. Those are wildcard files.
The wildcards are just random snippets of prompts that you can use in your own prompt text to get different results. My prompt pack has a colors file because I don't always want people wearing the same two or three colors of shirt or whatever.
There are lots of wildcard packs out there that will give you everything from clothing to whole scenes and all kinds of things. "wildcards" is a category in the models section of civitai, so you can do a search and browse through what's there. I'd recommend starting with the big pack from cyberdelia because it has all kinds of stuff inside. I've used those a lot in the past.
https://civitai.com/models/359653/cyberrealistic-humongous-wildcards-pack
Try playing around with them with images first because you'll get much faster results and you can see how they work. Then before long you'll be making your own to save you from re-typing your favorite scenes and options.
@darkroast175696 thank you for the reply, i appreciate the effort!
Oh thanks so much for sharing those! You rock!
Tip for those that are loading the goat's amazing workflows... most of them are NOT prefixed with __darkroast like the examples here so placing them in the 'darkroast' subfolder of 'wildcards' will NOT work! (place his archive at the root of 'wildcards') Note that you must reboot Comfy for AdaptivePrompts to reconnect to its folders if you shift things around)
Many thanks