1000 Prompts in 1 Click - Dynamic Prompt Wildcards for Automatic 1111

how to create hundreds of prompt variations with just one click hello my friends how are you doing this video is packed with information I'm going to show you how to use AI to create keyword lists how to randomize these keywords how to stabilize the output in automatic 1111 so much cool stuff let's get started so of course the first thing we need here is the extension this is called SD Dynamic prompts by the way big shout out to Don bumby for helping me with all of this so we go over to stable diffusion you want to go here to extensions and there you want to click on available you have here the load from button click on that to load this list and in there you're searching for SD Dynamic prompts when you found that on the right side you click on the install button and wait for the install to finish should only take a couple of seconds of course next as always you want to go to installed check for updates here so so click on that button and then you want to next click on apply and restart UI that is for the install and now when you go over for example to text to image but also for image to image and scroll down you will see you have your new area called Dynamic prompts of course usually this is closed up so you only have this short line and when you open it up there's a lot of settings in here a lot of things you can do I would highly suggest to you that you go to the link that I provide here with the tutorial and there is a lot of text in here on how to use this specifically because you can do so many things with this but of course I want to give you some really solid advice on how to get the most out of this wonderful extension so the first thing I want to suggest to you is how to actually build the list of the keywords you want to use here and for that of course we're going to use chat GPT so here I'm simply writing make a list of 25 female email haircuts you have these lists here and what you do with them is to Simply put them into a text file and you can see that every of these keywords here isn't an individual line so this is how they are separated but you can also see here that the keywords can be made out of several words so you can even write something longer for example describing a location or other things so you can really build here your individual lists now this is saved as a normal text file and then this is going to be put in the automatic 1111 folder into the extensions folder in there into the SD Dynamic prompts folder and in there into the wildcards folder down here now when you put these text lists in here is very important for you that you restart automatic 1111 and actually I already prepared a folder for you with various lists in here for example clothing for male and female hair colors hairstyles to traditional clothing styles things like that so now you know how to build these word lists with AI and how to save them inside of automatic 1111 but it's actually amazingly simple to use them you write your normal prompt and then where you want to have them for example here I'm having a list for hair color of colored hair not natural colors and then I also have haircut types so you want to have your underscore underscore then you have the name of the text file without the dot txt at the end and then again underscore underscore and you do this for every Wild Card list that you want to use so in this case we are choosing a hair color and then we also choosing a hair type now if you want this to choose it randomly from the list what you want to do is to put that into the prompt and then set the batch size down here for example sample to 4 and as you can see here from my result I'm getting a different hair cut and a different hair color every single time but of course even though we are using the same seat because the prompt is changing also the rest of the image is changing so that we have here some change in the clothing too not just in the hair color and the hair type I will show you methods on how to get more control over the output in a second so first I want to show you a second way on how to use this for prompting so instead of having these lists on your drive you can of course also write them inside of the prompt by hand you don't have to prepare any files so in this case I'm using an angle bracket or Chevron as I think it's also called and then I write a color have a straight line down write another color without a space in between and then another straight line down so you can put as many terms in here as you want font and you can put this at any point in your prompt for example for the light situation for the clothing for the H for the hair color for glasses for the type of shirt any kind of thing you want of course at some points you also want to choose multiple of these words at the same time in that case what you can do is that you will have here the number of how many you want to choose for example two and then you write dollar dollar sign and this will choose now two terms from this list inside of the angle brackets now in the next example that I want to show you here I have made two of these brackets so the first one has red green blue and pink as the color the second one has short hair bob cut and curly hair so different hairstyles that I've written here again in these angle brackets now instead of using the batch size down here what I'm doing is to set the batch back to one and I'm scrolling down here and you can see that I have a choice here for example of combination generation now when I click and activate this and I leave the value at zero this means it will create any possible combination of the words in here in the bracket so for example it's going to create red short hair red bob cut red curly hair and after that green short hair green bob cut green curly hair so you go and run through all of these Generations but you can also see here that you have a number of Maximum generation so if this is too much you can limit this with this slider here to how many pictures you want to have and below that you can also set the number of batches you want to have now the batches of course they are useful when you're using different seats now if you're using this combination generation in here you can set here the starting seat when you scroll down here under Advanced options you have here the choice to fix seed in that case that means that any seed you put up here will be used in the same way for all of the pictures now before I show you another way on how to get even more playful with your prompt variety I want to show you how to fix the output result so my first advice to you would be to use open pose that means I'm generating images here until a pose is found the composition is found that I like or you can also use a composition or pose from a photo from an image from the internet and then of course you scroll down here you open up your control net 1.1 I also have a really cool video about control and 1.1 that you should check out and it also has a trick in there for really massively improved quality of your images so what you want to do here is of course you load the image you want to use I enable my control net here I set my preprocessor in this case to open post full and the model to control net version 1.1 open pose now in this case because I'm setting my open post to full what this is doing for me it generates a stick figure for the body a point mask for the face expression in face position and then also tracking the hand and finger positions so this gives me a high control over the output now here I want to show you another method also where you change the hair color but not the clothing style for that we are going to use in pane so that means when you're in text to image you choose the image you like and then you click on send to in paint when you are in in paint you mask out the area that you want to replace then you scroll down here you choose only Mast you set up the amount of steps you want to use in this case I'm setting my denoise strength to 0.5 and I'm also again using control net but in this case what I'm doing here is that I'm using the input image that I have generated I'm setting my pre-process accessor to normal Bae and then my model to control version 1.1 normal Bae what this is going to do is that it creates for me a normal map and as you can see here this also includes the hair styles so that this is staying rather consistent with all of that in place I'm removing the variety or the Wild Card list that is about the hair cuts from my prompts so it's only about the hair color in here and then I click on generate and that should result in an image where the hairstyle is the same but the hair color has changed and because we are only in painting the rest of the image like the background and also the clothing style should stay the same now of course this extension has some more to it than just using your prompt list you can also have another automatically created variety for you and this can help you discover cool ideas and cool image looks that you wouldn't have thought of yourself so for that we're going back to text to image in this case you don't need these lists or these variables and you can actually make the prompt rather short so in that case for example you could remove all of this so you just have the starting point of the prompt and then you scroll down here to your Dynamic prompts now here we're turning off combination generation because we are no longer using that instead we are using the prompt magic you can turn this on here and you can actually choose down here the method you want to use so experiment with this so this is basically adding new prompt ideas to the end of your prompt based on what you have already written in your prompt so far you can also see with the prompt magic that at the end of the list you have don't apply to negative prompts on top of that you can also activate other methods so while magic prompt is turned on you can use I feel lucky and this is actually loading prompts from lexicon art that are fitting to what you want to create or down here you can turn on attention grabber this is going to use weights to emphasize different parts of the prompts if you have also here sliders for the minimum attention the maximum attention so experiment with that of course all of that prompt information is always saved with the image and in case you didn't know that automatic 1111 is saving the prompt information in all of your settings inside of the image so next time you want to recall on how an image was created you go to the PNG info tab you load the image in here in that field and then on the right side you have all of the prompts and all of the settings that have been used including the model information what kind of checkpoint model you have used up here and with these buttons you can send this to text to image to image to image in paint and also extra leave a like if you enjoyed this video really helps my channel and see you soon bye oh you're still here so uh This is the End screen there's other stuff you can watch like this or that's really cool and yeah I hope I see you soon leave a like if you haven't yet and well um yeah

Channel: Olivio Sarikas

Views: 138,494

Keywords: oliviosarikas, olivio sarikas, olivio tutorials, Adobe Alternative, Automatic 1111, Dynamic Prompt, Dynamic Prompt Extension, Random Prompt tool, Wildcard Prompt, magic prompt, automatic prompt generation, ChatGPT Prompt List, ChatGPT Prompt, Prompt Trick, Best Prompts, Prompt guide, Stable Diffusion, Generative AI, Ai Art

Id: bQK5diN59NA

Length: 12min 21sec (741 seconds)

Published: Mon May 01 2023