NEW Midjourney style Image Prompting in Fooocus! | RIP Midjourney

do you want to make AI generated images but you're not very good with computers maybe even the absolute Basics confuse you so software such as comfy UI only gives you a headache well fear not young wannabe nerdling for now Fus has IP adapter support to give you mid Journey like image prompting yes hello and welcome to More nerdy radent geekery where today we delve into the world of using the brand new image prompts avail ailable in fucus as well as being really easy to install even for Microsoft Windows beginners fucus has a simple interface allowing you to just focus on prompting no more spaghetti nodes or different tabs all over the place giving you all sorts of different interfaces no buus just has everything all in one place on the one screen thanks to these image prompts you don't really need to make lauras anymore maybe you have a person you want in a different pose no problem pick that person select the pose a voila person in pose the documentation gives us a quick overview of the features and also tells us this technically this feature is based on a mixture of Ip adapter a precomputed negative embedding from the fukus team an attention hacking algorithm also from the fukus team and an Adaptive balancing waiting algorithm as well okay so I guess now we know how it works but why did they do it oh it says there as well the motivation of these efforts is to achieve a best match to the mid Journey image prompt I think that's at least the fifth time mid Journey has died this year rest in peace mid Journey all right then this table gives us a little selection of differences between mid Journey some other stable diffusion image generators and fucus it it's mostly accurate but as anyone who has used IP adapter before knows many of the issues they're saying here really only apply if you use IP adapter on full power for example IP adapter does work with text promps in come for UI and you are able to get diversity just don't use it at full power Fus has exactly the same problem if you do that so here I've got my input image of a rather cool rodent stop at one and a weight of one I'm using the papercraft flat paper cut style there and do these images come out in that style well I'll leave it up to you to decide however if you lower the power and the default on this is actually 0.6 and 0.4 so this is exactly the same thing just with the default weights then I think you'll see that the style has more accurately been applied and that's much like in my previous video where I made the paper cut style rodent in comfy UI with IP adapter and found that it worked best around .3 so maybe take that table with a pinch of salt but the rest of the document is good in its simplest form here you can use one image without any prompting at all and you'll get a reimagining of that image of course you can use prompts as well there we've got a single image with a text prompt and we are turning that dog into a hat obviously what you should be doing is turning them into rodents but never mind you can have multiple images with all without text prompts as well so put as many images as you like in there and they'll all get mixed in together and you can mix Styles and whatever else you want so you know a big merry bag of stuff as it shows there we're mixing in a sort of little bird with a cute dog we're turning it into a cat and they've got Styles and each one of those different aspect ects will go into your final generation and as if that isn't enough there's also an advanced section as well which basically provides two different control Nets py cany and cpds all right I think it's much easier if I show you all these things so let's take a look at some examples when you start focus it will look like this and in order to use this brand new image prompt feature you're going to need to click input image and then select the image prompt tab there you'll get the four input boxes now the very easiest thing to do here is just to put an image like it says drop image here or click to upload so we'll put an image in there and then just click generate wait a few seconds and you'll have your final results based on your input image there you go not too bad or perhaps you do want to use a prompt Maybe maybe you want to add wearing a hat so that she's wearing a hat not a problem when you regenerate again you'll get some more similar outputs to the input but this time wearing a hat because that's what you've prompted for you can also apply multiple images and different styles now what I've done for these tests as I'm using this image prompt feature is I've gone over to Advanced and then on the style section I've unticked the default which is usually fucus V2 and slightly cinematic that's because I want to see exactly how much it is influenced by these images without any Styles going on in there so okay let's close these Advanced things down and there we have you know the bearded guy and the woman we've got the brownness in the background there so you can see it's picked out some of the different aspects but one thing that is quite evident is that it's a woman there's there's no guy in there and she hasn't got a beard so does it does it make any difference which order you put these images in uh no is the answer I came to it doesn't because I can swap those round and it will still be a woman what does appear to make a difference however is the image itself so some things are more powerful than others in this example here woman beats beard for example but if we try Woman versus rodent in a paper cut style as you can see that has given us rather a weird mix that's kind of 50/50 on both of them I think there all right let's try some other ones what about Woman versus famous painting as you can see there famous painting guy in a puffy shirt playing what looks like a guitar and it's still kept woman that's it seems to be more powerful than painting as well we've got the photo style but it has added that guitar in so some aspects just seem to blend really well with the image others are a lot more powerful perhaps you want different styles not a problem all you have to do is prompt for it or select from one of those styles on the right hand side here you can see mixing the woman with a picture of lots of shoes and the prompt watercolor painting of a woman in a shoe shop and there you go you see it's got the flowery top it's even got his apron all the different shoes in the background so it's kind of mixed the two images together and blended it with our prompt as well to give us those final Generations if you want to go really Advanced then down at the bottom there tick the advanced button and you'll have a whole array of different controls open up now images don't have to be image prompts as you can see there each image actually has three options it can be image prompt py canny or cpds canny being an edge outline and C cpds standing for contrast preserving decolorization which is quite a funny name as they only use the structure part of the images and it doesn't really do any decolorization all right probably best if you see exactly what this does so here we've got that picture of our woman again and we've got another picture of a woman but she is pointing up and we using that cpds and as you can probably guess there you go it's done the woman in that pose so it's a little bit more like the depthmap model it was originally trained from pyra canny does much the same thing so here it is exactly the same but this time it's pyro canny and once again the woman is emulating that pose mixing everything together multiple images control Nets prompts and styles can help you to give a little bit more control over your image Generations so there we've got bearded guy I'm using pyra canny and cpds for that pose and another image prompt of a bearded guy along with this prompt painting of a bearded hipster wearing jeans with an alien invasion happening in the background and there you go that's exactly what we get so he's in that pose we've got his beard we've got his little hairstyle we've got the Alien Invasion going on in the background and all these different art styles mixed in as well how much fun is that so why not give it a go today or you could just watch some more nerdy roaden videos

Channel: Nerdy Rodent

Views: 7,739

Keywords: nerdy rodent, ai, ai art, best ai tutorials, Midjourney for free, Ipadapter fooocus, Ai art guide, Sdxl image prompts, Midjourney image prompts, Midjourney free alternative, Midjourney for beginners, Sdxl for beginners, Ai art for beginners, Generative ai for beginners, Ai powered art

Id: 8mTWVMtq7zQ

Length: 9min 40sec (580 seconds)

Published: Sun Oct 08 2023