AMAZING!! You Can Create This AI Image Concept With Stable Diffusion Only

hi what's up everyone in this video I will show you how to make an outstanding fantasy image with a girl and dragon concept like this I'll show you the tutorial in a short and concise way so you can develop it yourself according to your imagination so there is no need to wait any longer let's get to tutorials before going into details there are a few things you need to prepare in the stable diffusion web UI that you are using first make sure stable diffusion has the licorice extension if you haven't installed this extension I'll show you how to install it open the LA chorus extension GitHub page you will be able to find information about lycorus as I explained before La chorus has a function similar to Laura but lycorus has a different deployment algorithm here we click and copy the lycorus repository link then open stable diffusion click the extensions tab select install from URL then paste the link there and click install when the installation is complete click installed then apply and restart the UI if successful you will get a like chorus menu next to the Laura tab like this foreign then the second thing we need is the after detailer extension on the GitHub page you can get detailed information about this extension after detailer will really help us when generating AI images in stable diffusion web UI this extension will help us make the face personality hands and other parts more detailed and precise we can use this extension in text to image image to image and also in painting scenarios 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 add a tailor extension like this the next thing we need is the Laura and lycris models the Laura models we need include lung China Dragon Laura concept low raw for lightning add Laura details and writing on a Laura model as for the lycris we only need the unlimited blade work lycris model all the download links for the Laura and lycurus models that I mentioned I have copied on a Google Document The Links of which I have placed in the description of this video you can download it from there easily without having to search for one by one on the civitai.com website all right if everything is ready now we open the stable diffusion web UI and I've prepared the first draft prompt for you guys to follow I have copied all the prompts that I used in this video on the Google Drive document you can copy it for you to use freely here we are going to include a prompt with a photorealistic concept a girl with golden hair with the addition of some composition Styles as you can see here first here we will only enter one Laura the Laura we need in this prompt is lung Dragon Laura to enter very easily we click show or hide extra Network then click the Laura tab and type in the Laura keywords that we will enter here we enter Laura Long Dragon then just click on the Laura we choose then the second Laura that we enter is ADD detail Laura this Lora function is to give a more detailed touch to the AI image that we will generate then for the parameter we set it to 0.8 for Laura's low Long Dragon while for add Laura's details we set it to 0.7 you can customize this parameter number yourself the higher the parameter number the greater the Laura image will be in the AI image later the settings that I use are like this stable diffusion checkpoint or the model that I use is the XX mix 9 realistic model you can use any other model you like but if you want to use a model like the one I use you can download it via the link that I have provided next I set the sdvae to Emma pruned84000 I set the clip skip to 2 then I set the sampling method to DPM plus plus sde Keras and set the sampling steps to 30. for the size I set it to 512 X 768 while I set the CFG scale to 8.5 then here I suggest you to turn on the hi-res fix option this option will refine the Laura details that we apply as well as upscale the base image to be higher by default hiresfix will upscale the image by upscale by two but for those of you who use a graphic card with vram below 8 GB I suggest changing the upscale by option to one and a half times then for the upscaler option you can choose whichever you like here I use 4X underscore foolhardy ramachry another important thing is the denoising strength as in upscaling an image to image we set the denoising strength at 0.2 to 0.35 here I set the denoising strength to 0.3 when everything is ready it's time for us to generate the generate process by activating hiresfix takes longer than generating a normal image this is because the image is generated in two steps but if you upscale it by one and a half times the process won't take too long okay and here are the results we get the image of a girl side by side with a dragon like this it's awesome image here we pay attention to the face when we use Laura with high parameters and turn on The Hires fix option we will get that the facial parts of the characters are sometimes not very precise like this this is where we need the add a tailor extension to activate it we just scroll and click the detailer extension tab then check enable add a tailor in the model at a tailor there are several choices of models that we can use there are face models hand models to person models here we just give an example using the face underscore yolov 8npt model in the description column we can enter simple prompts such as beauty girl natural girl Russian girl and so on or we can also combine it with a textual inversion face Laura face or like chorus face model here we just fill in a simple prompt Russian girls Gal Gadot then we generate [Music] and here is the result now part of the face becomes more detailed and precise even though here we can't get gal gadot's face because the face model doesn't have a Gal Gadot face database we can overcome this by using textual inversion or Laura model face and this is an example of an image that I have generated with a combination of the Laura Emma Watson face model Laura on the Ada Taylor extension her girl face became similar to Emma Watson then because the parameters we enter are small the level of similarity is not maximal okay we'll show you the next image concept here I have prepared The Prompt I wrote The Prompt with the concept of woman holding sword with dragon long Aura Spirit to create the sword holding image with precision and accuracy we need lycorus model unlimited blade works here we can see the lycorus model that I have installed with parameters 0.9 then I added Laura Laura to make the image with a darker color setting I don't need to explain the model image settings and the use of the add a tailor extension anymore because I have discussed it in the first prompt but if you want to see the settings that I use then you can set them with a composition like this okay if everything is ready now we generate the image and this is the result we get the image of a woman with blade and there's a dragon long Spirit behind it to get an image with a combination of multiple Laura and lycris like this sometimes we need to generate it several times because in other image results sometimes we get only girl images only dragons or a combination of both to find the right combination we need to play the numbers on the Laura or lycorus model parameters here I try to derive one point from the figure for each Laura and licorus model and the result is like this in my opinion the concept of a girl holding weapon with Dragon Spirit looks awesome the combination of prompt and Lora gives a dramatic impression with multi-colors like this okay now we come to the last concept I still use the same basic prompt as before but this time we want to make the girl ride the dragon for that we need to install the Laura riding a concept in the prompt we can see I've entered the Laura with the parameter number 0.6 this is intended so that the generated AI image can later be combined with other Laura with more Precision okay now we generate and here we go here are the results we can get an awesome image with the concept of a woman holding a weapon and riding a dragon that is dramatic like this but there is a small detail that we need to pay attention to the facial features of the characters still look ugly and imprecise this still happens even though we have used the add a tailor extension to fix the result we need to do manual upscaling on image 2 image I have practiced this method many times and done it in previous tutorial videos we select an image then send to image to image in the settings we set the method aside to DPM plus plus sde Keras then select resize buy and set it to size one and a half I set the denoising strength to 0.33 like this then also enable the detailer extension enter the same prompt we used for text to image then scroll to the control net Tab and enable it select tile in the control type and leave the preprocessor and model at default don't forget to select control net is more important then scroll to the scripts tab select ultimate SD upscale and in the Target size type we set it to scale from image size for the upscaler model just decide according to your preferences here I use 4 x Ultra sharp furthermore when finished we generate [Music] and here it is the result now the girl's face is beautiful and precise like this then the dragon object also becomes more detailed plus the image has a bigger resolution with high quality I've also upscaled other images and this is the result this is really awesome image what do you think about the image that I generated this time write your opinion in the comments column that's all my video this time I hope it's useful and you like it thank you for watching and see you in our next video

Channel: SahiNama

Views: 8,091

Keywords: ai art, artificial intelligence, midjourney ai, stable diffusion, ai art generator, how to use stable diffusion, how to use ai, how to train stable diffusion, stable diffusion training style, stable diffusion best tutorial for prompts, stable diffusion best tutorial, stable diffusion negative prompt, dragon girl, openpose controlnet, open pose tutorial, lora tutorial, lora stable diffusion, upscaling stable diffusion, upscaling stable diffusion images

Id: o4SDPUEO17Q

Length: 11min 15sec (675 seconds)

Published: Sun Aug 06 2023