0:00 welcome to episode 6 of our comfy UI tutorial series I've created a file with over 300 art styles and in this video I'll show you how to use it in comfy UI this file was originally made for Forge UI but I found a way to use those automatic 1111 and Forge art styles within comfy UI I included a range of styles from painting and illustration to Vector photographic and a few experimental ones that you can enjoy experimenting with the file I mentioned can be downloaded from go Google Drive I've included a link in the description it's a CSV file that lists the name of each style along with the positive and negative prompts you can click the download button to save the file to your computer then place it in your comfy UI
0:45 folder to confirm it's in the right location make sure you see the main. Pui file in that folder I will use a basic sdxl workflow that you can download from my Discord server in the AI resources Channel Once you drag it into your interface it should look similar to this one just like we did in previous episodes but before we load that file let me first explain how to combine two text prompts into one so that when we load the file it will make more sense I'll move the nodes around to create some space I will copy and paste the positive prompt node using the control+ C and control+ V shortcuts I'm splitting the prompt into two parts leaving the first part in the First node and the second part in the second node as you can see if I try to connect the second node to the positive prompt it will
1:31 disconnect the first one but I want both of them to be used together as one I'll use a node called conditioning conat for that you can search for concat in the node list concat is short for concatenation which means joining multiple pieces of information together now connect the first node to the conditioning to input and connect the second node to the conditioning from input this way both text prompts are combined in this node concatenating two text promps is like mixing two flavors of ice cream to create a new unique flavor now we can connect this node to the K sampler positive input if I try to cue The Prompt and get a notification about missing input on clip it means I forgot to connect the clip node to the low checkpoint clip I'll connect that
2:18 and now it should run just fine producing a robot image from the combination of those two text prompts I can replace the robot with something else like a bunny and I'll get an image like this we can use more than two positive promps if we want I'll make space for another node and copy and paste the second node don't forget to link the clip inputs now I'll split the prompts into three parts you can split them however you like but I'll do it like this the first node will be the subject the second node will contain details about the subject and the third node will have information about the environment there are additional nodes and settings that allow you to control how much each prompt influences the result but I'll cover those in a future episode if I try to connect the third node it disconnects the second one so
3:04 what can we do we can add another conditioning concat node I'll make some space for it now the first and second prompts are concatenated in this node and become one you can connect that to the first input then connect the third node to the second input finally connect everything to the positive prompt so the information from all three text encoders flows like Rivers merging into a single stream that goes to the K sampler if I cue The Prompt I get a robotic bunny now you can easily change the first prompt to another animal if you want however since the prompt is only one word it doesn't have as much influence as the robot in this case for example if I try with bird the result is still more of a robot than a bird but if I add more
3:50 details like a red bird with wings I get something closer to what I was hoping for this was just to demonstrate a simplified version of what comes next I'm deleting the third node because we only need two for the styles to work now if we search for a node to load Styles or CSV files we don't have one the one from here is for something else I searched for a solution for a few days and eventually found a custom node that allows us to do that click on the manager then go to the custom nodes manager at the top this will let you install a custom node from the internet similar to adding an extension that allows for extra functionality search for load Styles and you should find the Styles CSV loader the description mentions it's useful for migrating styles from automatic 1111 click the
4:35 install button and wait for the installation to complete after that press the restart button and a new comfy UI tab will open with the updated interface Now search for the node load Styles and you'll see the new node called load Styles CSV that you can select if you put the Styles CSV file in the correct folder all the art styles from that file should appear here you can scroll through the list using the mouse wheel and use the top search bar to find a specific style I'll leave it with no styles for now so you can see the effect once we select a style from the list the CSV file contains just prompts which is why I showed you how to combine prompts from different nodes essentially we now want to combine the positive prompt with the art styles
5:23 found in the Styles CSV file when I try to connect the positive prompt from Styles it doesn't let me connect to the K sampers positive prompt or to the clip input of the prompt so what do we do it took me a while to figure it out but here's the solution right click on the second positive prompt node go to convert widget to input and select convert text to input we use this function in a previous episode with a primitive node but now it lets you add the text from styles to the text input of the text encoder I'm changing the prompt to something like a cute bunny in the forest then I'll arrange the nodes a bit and test the workflow um without an art style using the Juggernaut X model we get something like this now if I select
6:08 an art style such as painting abstract expressionism and cue The Prompt we get a painting like this you can also search for styles for example if I search for cute select Character cartoon cute and test it I get this cute bunny we have the positive prompt we added the positive prompt from Styles and the negative prompt but the Styles also include a a negative prompt for each Style so let's see how we can add that uh I'm duplicating the negative prompt node then right clicking to choose convert widget to input and selecting convert text to input I'll make it a bit smaller then I can connect the negative uh prompt to the clip input but now we need another conditioning concat node connect the negative prompt into this node along
6:53 with the negative prompt from the Styles then link the output of this node to the K sampler node so now we have loaded the Styles CSV file into this node and that file includes both positive and negative prompts for each style um we have uh the positive and negative prompts like in any workflow the positive prompt uh combined with the Positive prompt from the Styles goes to the K sampler the same process applies to the negative prompt we combine them and and send them to the case sampler if I run this workflow I get this cute little bunny now that we have a working workflow I prefer to simplify the interface to make it more compact to do this I select all the
7:38 nodes except the save image node using control then rightclick and choose convert to group node I name it something like text to image you know look how cool and simple it looks now it doesn't seem so overwhelming anymore I'll resize it to fit better on my screen let's see how we can improve this uh interface even further if I rightclick on uh this group node there's an option called manage group node and from here you can move things around for example I can move the prompt up and when I save it the node is updated you can also hide elements and rename them I'll try to make it similar to the automatic Evan 11 and Forge interfaces from the top we can load the sdxl model then we have the positive prompt and the negative prompt it currently says clip
8:23 text encode so let's change that using manage group node from there you can rename it from the widget options and save it now it will say negative prompt next we have the styles that you can select from the list or search for by typing for example if I select anime chibi illustration I get something like this I added some extra Styles called experimental uh for example if I test a Legends digital painting style I get a nice bunny character similar to game art you can get even more creative with Styles like biomechanical Angel it's fun to experiment so try different prompts to see what you come up with I organized them them by categories so if you search for photo you can see all the photography Styles if you search for illustration you'll find different
9:10 illustration Styles it also includes more categories like design and Vector which cover flat and silhouette Styles crafts is as great for inspiration if you're a crafter or for generating images that resemble Craft products from Ceramics to sculptures from different cultures to origami other categories include painting fashion art 3d and experimental if I select the cute cyber ninja style I get this cute 3D character if you want to avoid being affected by any art style just choose no style uh you can save the workflow and give it a name that makes sense for you I'm saving and uploading all the workflows from my episodes to the Discord server and the easiest way to access the workflows is
9:56 to click on the link from my YouTube channel and accept the invite I've created a news and updates Channel and moved all the workflows into the pixaroma workflows channel each entry includes the episode number a description of what it does a preview and the files you need to work with just download the Json file and drag it into your comf uh check other channels like comfy UI share your art or see what others have shared you can also share prompts and explore more AI resources back to the workflow let's convert it back to nodes and drag the save image node into view uh now I want to show you how to combine two art styles since this will make things more complicated I'll change the colors of the nodes green for the positive prompts and red for the
10:41 negative prompts I'll also update the title of each node to make it easier to identify which prompt comes from the text and which comes from the uh Styles uh for the load Styles note I'll use a yellow color and change the title to include number one so I can duplicate it and have another one that includes number two since I don't have much experience with nodes it's took me a while to uh to figure out how to connect them uh I've created some space between the positive and negative nodes I duplicated the positive prompt node the one used by the first Styles node and connected it to positive prompt and the clip then I repeated the process for the negative prompt duplicated the negative prompt node and connected it to the negative prompt and the clip here I had a bit of a brain freeze and wasn't sure
11:26 which connections to make then I thought let's take it logically we have the positive text prompt the one we added and the one from the first style we just need to combine it with the second one from the style so I duplicated the conditioning concat node I connected the first condition on top moved it a bit for clarity and then connected it to the case sampler for the second input I use the positive prompt from the second Styles node I fix the positive prompt and let's do the same for the negative prompt if this feels confusing just download the workflow it will get easy iier over time the more you play with nodes the more comfortable you'll become so connect the first conditioning concat node to the first input and the second negative prompt
12:13 from the art style to the second input then link everything together the case sampler uh so this is how it looks arranged and you can get it from Discord I have a robot in a cave prompt I'll select a style like Chinese ink brush and use a fixed prompt so we can see the effect it has if I run the workflow I get a painting like this now let's deactivate the first style and add a graffiti style for the second one um running the workflow gives us something like this when both Chinese ink graffiti styles are combined the result uh looks like this if you're wondering what happens if you reverse the order of the Styles the result is almost identical just like we did with the first workflow you know you can select all the nodes
12:58 and convert them to a group node then adjust settings uh to look how you want I've already done that and saved it as a workflow after you download the workflow you can simply click load and select the compact version that looks nice and clean like here by default no style is active so if you run the workflow it will look something like this um I selected the long exposure style and when running the prompt robot in the cave the robot got lost uh due to the strong effect of the long exposure art style to balance it out I added a few few more words to The Prompt if I select a second style like fantasy portrait I get a glowing effect keep in mind that some styles are stronger than others and can overpower them and for example with
13:44 the style flower photography I didn't get any flowers because the robot the cave and the long exposure style were too dominant simple fix is to add flowers to The Prompt uh now when I test it I get this nice scene with flowers uh in the cave to achieve the desired effect uh try combining styles from the same category or similar categories such as photo with fashion or crafts Vector with illustration or painting with art for Vector silhouette combined with logo I get a nicely styled Vector robot and if I change the second style to chibi the result is a chibi looking image now I can change the first style to something like manuscript and the second one to
14:31 cartoon character and continue experimenting with different styles uh for example selecting the Avatar Style gives me these cute little robots you can also play with with colors or introduce other elements into the scene to further customize your results as you can see these Styles can be quite useful once you learn how to connect the load Styles node properly you can use it in different workflows with various models you can edit the styles using the text editor like notepad or note pad Plus+ but make a copy first in case something goes wrong or just redownload the Styles file and replace it if you messed up the file I struggled for a few hours with errors just because I manually deleted some Styles instead of adding and
15:16 deleting manually consider using tools like automatic 1111 style editor or Forge UI style editor if you do edit manually it's best to modify an uh an existing prompt that you don't use focusing only on the text between the quotes that's all for today thank you for watching if you found this episode helpful please leave a like or a comment have a great day [Music]
