How To Get The Most Out Of Playground AI Filters

well hello good people today I wanted to talk to you about how to utilize the filters within playground Ai and how to really take advantage of them one of the things that I often do is start with the playground V1 model as you see here I've already generated an image and then I will change it back to stable diffusion 1.5 bring it into image to image and perhaps use a filter like Polaroid for this demo I'm just going to show four random Generations not utilizing the seat and now you see we have four different variations of this original image that we did in playground V1 this is a great way just to get a different variation of your particular image now let's backtrack a little bit let's start with the models you can think of these models as your foundation this is where the original data sets were trained on 1.5 is just a bit more versatile than let's say 2.1 or as with stable diffusion 2.1 it's a bit limited based on how it was trained with playground V1 think of it as another data set that was trained on top of stable diffusion 1.5 I do encourage you to use playground V1 because the way it's been trained it tends to have a better structure for certain scenarios like character hands for example and of course because it's got a higher dynamic range the colors are typically more contrasty more vibrant and overall it makes a great foundational model to start with I'm going to quickly demonstrate the differences between the models so here's stable diffusion 1.5 this is without any filters it's just raw stable diffusion 1.5 now let's take a look at stable diffusion 2.1 again with no filters and here's 2.1 and it looks great and then we'll do one for play ground V1 so all very different looks I encourage you to experiment with them all now let's talk about the filters let's open up the filter panel here and you're going to notice that in regards to playground V1 only these filters are available and reason being is because these filters are basically pre-made text prompts in the early days of playground there weren't any dream Booth models at the time so I think of these filters like presets these are prompts based on specific looks to add on top of your prompt if you look at this prompt for this car we're still using playground V1 The Prompt ends at inspired by the movie Tron I'm going to go ahead and select Lush illumination we'll generate a similar image without using the seed and here's the image if we go into my profile gallery and if we look at the end of The Prompt here you'll see these additional words were added to the pro prompt so for this particular image there were more prompt words used to get this particular look so these 18 filters are basically pre-made text prompts and the reason why I bring this up there are certain words on these filters that you don't have to put in your prompt I have a spreadsheet here that I'm going to link in the description below but as you can see a lot of these words are used fairly repetitively for example delicate detail you see words like sharp Focus Studio photo intricate details highly detailed you also see Sharp Focus here in radiant symmetry different artist Styles so knowing these particular words you don't have to put them in your prompt as they're already there one of my favorite filters is neon Mecca what gives that filter its look are words like neon Ambience gear Mecca grun inch abstract black oil the other great thing about knowing what words are being used for these filters is that you learn about that particular look and how it can apply to your image now we're going to switch this to stable diffusion 1.5 go back into the filters and now you're going to see that we have more filters available for stable diffusion 1.5 not only do we have the previous 18 but we also have these ones like colourpop and support play tune so on and so forth all the way down to this one geometry the style these 23 filters are known as dream Booth models you can think of dream Booth models as an additional data set trained on top of stable diffusion but these are not text based these are trained on a specific style for example Polaroid or analog diffusion these models are trained to look like film photography be they sort of have that vintage look especially if you're doing portraits and these dream Booth filters are only available with stable diffusion 1.5 one of the main differences is that to trigger the model they have these trigger words when you fill in your prompt and you select something like colourpop style this trigger word will be placed in the beginning of your prompt to get that particular style now if you look at examples like play tune or even poly mode there were some additional words to further enhance the style but for the most part the main model is triggered by these words in bold as we scroll down to some of these other models you see there are no supporting prompts just the trigger word itself so I highly encourage you to experiment with the different filters in these examples I was just messing around with some song lyrics the song Moon River and I use warm box for this and you see it's got a really flat palette an acrylic painting type of style this is the same prompt using radiant Symmetry and this one using neon Mecca and the great thing about the original text filters is that you can take the words from it in either add more to it or take some things out if certain things aren't working you don't necessarily have to select the filter now I can take the same image and let's say I wanted to change the style of it I have it going into image to image here with playground V1 and the reason why I'm bringing it into V1 is to just give it a bit more structure and I do want it to change a little bit so here are the images side by side this one was the original and this one was generated using the playground V1 model so in terms of details you see that it's got more details in the mountains and the clouds in the sky it even threw in a couple people here on a canoe I guess then of course we can repeat the same process until we were happy with the end result and with this example I started in playground V1 I wanted a little bit of an anime look just to get the initial idea and then I brought it into both Dark comic and storybook to end up with this image so it may take some experimenting but it's super fun to figure out certain looks that you can use and you'll start to be more familiar with certain filters that work best with certain situations for example for portraits colourpop instaport analog diffusion Polaroid perfume all of these work very well for portraits if you're looking for more of a cartoon anime style retro anime play tune retrofuturism even Dark comic and storybook work well for those types of images for car photography I like to use Polaroid analog diffusion even dream Haven but the key thing here is understanding the look and style of the filter and how it can be implemented into your image now I hope you have a better understanding of how the filters work and how you can utilize them now make sure to stay tuned as we're gonna go really deep into canvas in the next videos to come there have been quite a few new additions to Canvas that we're going to go over until then my friends this is playgrounds AI

Channel: Playground AI

Views: 9,837

Keywords: How To Get The Most Out Of Playground AI Filters, playground ai, playground ai filters, playground ai tutorial, stable diffusion, online image generator, free stable diffusion, ai art, ai image generator free, ai image generator, text to image ai

Id: W3AqBrPmBsI

Length: 8min 36sec (516 seconds)

Published: Mon May 08 2023