HAVE FUN!! Turn Your Favorite Game & Anime Character Into Realistic With Stable Diffusion

hi what's up everyone in this video I will show you how to make a 2d character image from your favorite game or anime it becomes a realistic image with stable diffusion into like this [Music] so no need to linger let's get to the tutorial actually the tutorial that I will give is not something new for users who have been using stable diffusion for a long time but maybe this will be useful for those of you who are new on learning to use stable diffusion web UI before discussing technically there are things we need to prepare the first is the addy Taylor extension this extension will help us make it easier to make the face of the character really realistic but this is optional you can use it or you don't need to use it for those of you who don't know how to install it I'll show you how but if you have installed it you can just skip to the next step to install it we open stable diffusion then click extension and install from URL then copy the repository link on the GitHub page and paste it in the URL tab then install it and wait for it to finish when finished click installed and restart and apply web UI then later you will see the added tailor extension like this [Music] next is to determine the checkpoint model that you want to use this is very important because it will affect the results of the image that we generate later for game characters anime and fantasy things I suggest you use a checkpoint hybrid model such as fantastic mix Photon and the like if you use a realistic model such as realistic Vision it will be difficult to get precise image results step is to prepare the basic input image you can search on Google image Pinterest deviantART and so on here I have prepared several valorant character images which I will make it realistic versions also make sure the image you download is a good quality image if possible also has sufficient resolution do not download images with low quality and resolution because it will significantly affect the results when the image has been collected now we open the stable diffusion web UI to change 2D characters to be realistic there are several ways that are often used by users we can use text to images with the help of control net and then you can also use image 2 image here I will show a simpler way using image 2 image we click image 2 image tab like this then drag and drop one of the images that we have prepared earlier like this I will try to make a realistic character Joy from valorent then in order for the image that we generate to be more precise we still need a prompt here I enter a prompt realistic photo of one girl wear glasses wear green beanie black hair Wear Yellow Jacket holding AK-47 prompt will really help AI to generate images more easily and precisely for that make sure the prompt that we enter has the characteristics that match the input image that we enter next we also need to enter a negative prompt I have copied the negative prompt that I usually use in image 2 image you can see in the description of this video to copy it next we go to the settings for the sampling method I use DPM plus plus sde Karis next for the size here we can enter the manual according to what we want or if we want to follow the basic resolution of the input image all we have to do is Click Auto detect size from image 2 image like this then later the size of the image that we will generate will be the same as the input image for sampling steps you are free to use whatever I usually use 20 to 30 sampling steps in image 2 image the next thing that is very important is the denoising strength for denoising strength we make sure not to use parameters that are too high because the higher the denoising strength number the more freedom AI will generate the image so the result will not be similar to the input image to get results similar to the input image I suggest you play at 0.5 and below this will limit the imagination of AI algorithm so he will generate images with more similarity and precision both in detail color and so on here we set the denoising strength to 0.4 like this next we will use the add a tailor extension but this is optional you can use it or not to use a detailer we need to click on the a detailer then on the a detailer model we set it to face underscore YOLO like this and at the prompt we enter the face of the character we want to generate here I enter one girl wear glasses in the attit Taylor we can also enter textual inversion faces or Laura face models it's very flexible you need adjust what you like in here don't forget to click enable a detailer like this and when finished now we generate and wait for the results okay here are the results we get a realistic character Joy from valorant like this everything looks similar and realistic from the clothes weapons held belts to the faces and skull caps that look realistic this is very easy and fun now we try with another character here is jet from valorent we drag and drop the image then customize the prompt with the character then we scroll down and also adjust the prompt in the detailer and we try to generate and here it is the result here the hand looks less precise this is because the input image also depicts the hand with a composition that is also not very detailed we can fix things like this in in painting section next we will try to combine this technique with Roop for you who don't know what is Roop and how to install it you need to see our previous video about the roof extension I have included the video link in the description of this video as before we drag and drop the input image like this then adjust the prompt to be more specific then scroll down and we deactivate the detailer because for the face we will change it more specifically with the Roop we scroll to the root Tab and we drag and drop the face image we want like this here I use ayaka miyoshi's face from the Alice and Borderland Netflix series then we enable the loop and when finished now we generate and here we go here it is the result we can find a yakamiyoshi face perfectly integrated into her character image it's fun because this way we can just insert our own face and we can cosplay into anything this way what I need to make a note for you in using image 2 image the denoising strength is the most important thing if you want to get precise results all you have to do is change the denoising strength number but remember the number must be below 0.5 it can be 0.4 0.3 and so on last trick for me in this video we can also combine this technique with the upscale ultimate SD so that the image results are more detailed I have also often discussed the SD upscale technique on this YouTube channel you can check the link in the description of this video by upscaling with SD upscale the image we can produce is like this this is an example of an image that I processed using ultimate SD upscale and here are some other images that I produced with image 2 images in stable diffusion so what do you think about this video and if you have input or corrections to my video please write it in the comments column that's all my video this time I hope it's useful and you like it thank you for watching to the end see you in the next video

Channel: SahiNama

Views: 12,738

Keywords: stable diffusion, ai art, stable diffusion tutorial, img2img stable diffusion, img2img, turn 2d image into realistic, make cartoon into realistic, img2img tutorial, change anime, anime in real life transitions, anime into realistic, stable diffusion tutorial for beginners, stable diffusion tutorial for video, stable diffusion tutorial for free

Id: f-UiCtJndqk

Length: 7min 59sec (479 seconds)

Published: Sun Aug 20 2023