AI Smooth Character Animation

hello there in this video we'll focus more on animation side SD and primarily we're going to do this nice and smooth animations based on the video so we can extract all motion pauses everything almost and even more we're going also have it like example here a few frames I extract it on different pauses for animation so we can take them same smooth beautiful animations flicker free with a face swap first an important step we need to have a stable diffusion installed on your local machine with aftermatic 1111 installation I will provide link down below for you so you can go check well we need to have it install control net available Roop and other extensions I am using the resolve studio and it's a free actually resolved DaVinci Resolve one version S3 you can download it links will be down below for this I extract frames and also wear a combined movies however you don't need to use this you can also just download a shutter encoder which is free applications and allowed you to extract frames and and put together techniques what we're going to use it will utilize some batch or script processing but we have it also other ways we can do it and I have another video linked down below where we're utilizing the batch processing so it's meaning you can have multiple frames you can download it and let your system process this way however the problem using batch processing it is does not retain information about the first frame or other frames for this we're going to use a special script that we're going to run allowed us to going over the images and process as a batch processing but only with specific value and apply effect of the first and last frames the script what we're going to use it frame video render for stabled Fusion to download we need go on this page again link below click on download now this is open donations if you feel generous You Can donating but it's not required but it does personal from my perspective I do like when people reward my work so if you think this script is helpful you even one dollar it's great donations let's go ahead next when you click you come to this page where you want to download it and you want to click and save this inside their folder it should be located stable diffusion script folder you can put it inside here enable the noise multiplier on the top and we've done this several times ago but if for some reason you forgot how to do this let me show you you want to go navigate in your stable diffusion folder go inside the model and open share it files share it python click and go open edit in a text pad or whatever you like it inside this shared when you open you want to search and find Initial noise multiplier and we want to set this to 0 0 by default it will be said I think 0.52 we just need to set this to zero zero so it's allowed us to control amount of the noise we needed after you're done save this and that's let's go back to our stable diffusion interface we'll go to inside the extensions so let's go inside the settings and enable some of our new scripts and other elements but a way to the read for the script to read you actually need to restart your server so it's meaning whatever Windows you have your server running just close it and run the batch file again to restart your server okay let's go start with the first very simple go and control net and then control net you can notice right here you can enable or disable do not append map to Output so it will just generate showing we don't need at this point you can check on this and be sure they allowed other script to control net enabled so this is almost all other extension that using need to access so this tunic check in all you notice right here I have a three um instance of control net enabled so I can use a three model control net at the same time Next Step we're going inside the stable diffusion Tab and be sure the email with image images do exactly amount of steps slider specified is enabled this is will help us to create more consistency result let's enable this multiplier slider for this we need to go back to the user interface scroll down where you can see the options the display on the top and currently it's SD model checkpoint right here we also have the multiplier and grid save for example if I disable this will remove it when we restart the interface to add what we needed noise multiplier you just need click on an empty field whereas a pop-up Windows you're going this expanded go find the initial underscore noise underscore multiplier and select after you select it it will be just added right here in line for you okay next jump to the top save image and grid and be sure they always save all generated images is enabled well usually this is required if you work with the form or any other ones animations so you should most likely it is enabled for you but if it's not just check and enable after this when you're done click apply all settings and reload UI you should see right here on the top noise multiplier up here in my case you remember I remove a checkbox for the Grid it's gone top down below in a script we need to go to image to image script let's scroll down and right here click on a script and you should have it multi-frame in your case it's all says better multiple frame b072 I have a few other versions and reason I was modifying python so just adjusting some script rewriting but generally yours you should have just one script and of course you should have a control net and Roop I'm currently will use it stable diffusion checkpoint delivery we needed to take our video and split on a segments we can do with other applications but for now we'll just go and use it shutter encoder select video drag in drop inside the windows that you want to split on the parts okay so next choose the function we'll click down and select to image creation jpeg you notice this is change the image sequence right here create sequence and of course we can change if we need it on our destinations not as I enable check and I put it in an area where I want it I would recommend for you to create specific folder and put it all animations frames there and for example right here I have it AI stuff frames so this is where they're going to place if you need to change the resolution you can click down and change the solution that you needed that also if your video card have a small amount of memory you may be one resize because you can resize images here as well as you're doing this make them a little bit smaller like 512 by 512 or some other resolution when you're done just click on start function and this is will take and you can see we'll process and extract its cabinet very very fast and now if I'm going inside I can see right here all of these frames was extracted and placed okay where we're done with installing all necessary components now we can start working directly with the images we're going to use the image to image type remember we created some of those frames let's go ahead and grab our first frame we'll drag and place it inside the image you need to have some image here even if it won't be affected but otherwise you will receive air Also let's go ahead and put it prompt so I'm going to adjust my width to match the image resolution we also will leave it as just the size crop so this is same sampling steps we can go to 25 just a little bit higher safety scale and the noise strings will leave it as default so on top noise multiplier if you have it a little bit higher are just bring down to the zero as well we don't need any noise at that step let's go ahead and click generate and here we have it our slimy nothing resemble with what we wanted but it's okay for this reason we're going to use the control net expand control net and you notice right here code one two and three same from the settings that I said before for this also to enable when you take a drug and drop one frame and just to be sure I'll just Place those frames in all three control nets for the future use first control net I want good to enable and we're going to use it so let's go enable this and this is preloaded model for this if you don't have it any model or you're missing them for some reason I will provide link for you down below in description where you can go ahead and download it all missing or just all of control net models when you're selecting picture picture you should have it no preprocessor because it will using image they put it in and we use it model we're not going going to use this image for everything usually when we process in control net we use the batch file where we can preload it multiple images but when we pre-load batch files we cannot actually reference one image to another this is what the magic of the script will be it will have a single image we put it with control net it's enabled if you have it about eight or less gigabyte video card enable low vram and also I want to enable Pixel Perfect this is will take resolution original image and readjust for the control net to calculate based on all size of this let's go down below we leave it control net as default value and let's set control net is more important settings here leave it everything else as default now we're going above and let's click on generate right now we have only one control net enable and it's kind of nice they have this highlights color showing me one is only work covered our image very nice model was excellent working and does not affect the backgrounds which are actually very nice and right here we have our alien sometimes we I've noticed with aliens your hands maybe does not look good sometimes maybe missing fingers and everything for this we're going to use it another model of control net we enable right here Pixel Perfect Sam what we did before only we're going to use it to open pause and I want to use a full open pose because in this case if we look on the pre-processing but right here you see this teeny tiny icon this will enable preview but notice we have a face hands and we have a fingers so it's kind of nice to keep it consistency in a limbs if we want to do this as well same control net is more important leave it here and let's put this down to 0.8 for the weight because we want override let's go click generate and you'll notice right here pause how it's connected the other post should be close but a little bit different so because it will be little bit more with a sink of current pause if you like it what you see you always can go down to the seed and click on this recycle icons so we'll reuse our seed that we have before so this is our two control Nets we're almost ready third one is very very optional if you go on this control net what are we going to use it it is controller called temporal Net TV and this is kind of will analyze several images and applied the noising in some case if you ever use a DaVinci Resolve with a the flickering this is on a lighter version of this deflectoring when it's used analysis and may remove some information so this is need to be very careful I usually if I used I used very low value for this preserve it but it is another ways to remove it again we'll use the enable Pixel Perfect control net is more important in this case and let's render one more time so this time we have it our model established we have three different control net and by the way this if you don't have it temporal net yet downloaded right here will provide for a link to this file they can go down and download it div control net SD temporal net 723 megabyte this top just download it and place it in your stable diffusion folder extensions as the web control net models inside here this is if you install the stable diffusion by using extension tab inside the aftermatic if you are installed using GitHub then you will have it in a control net directly models and control net you will place inside here you don't need to start interface if you want to read just click on this reload button which will refresh and read all models inside the folders and then you can select it let's go up and check and you can see just little bit clean up but model look very nice so when you like how it's look play a little bit more maybe with a prompt negative prompt as you need it but I think this is work okay for me so next I want to create animations usually we can run through the batch to process all of them but we will have it little bit inconsistency if we do this so to do more consistent result we want to use the script for this let's go ahead click down and you want to select the better multi-frame video render 0.72 now notice we open several properties options important for us to be sure we're using the noise string 7.7 plus same we have it here we'll set seven five next this is important to reference of what images that should be rendered and it's referenced by default you have it first generated guide image and historical if you need it we won't go through all of them for this reason we just will use the first generated image as a reference and when I loot back what is happening after render it's going pre-loaded image and based on this same we need to use our first generated image as a loop back Source the next we need to upload the frames that we're going to use because it is need to read before we have a batch files now this will work as a batch file please click on upload guide frames navigate to the place where we extract frames and select all of the frames that you want to process I have it currently 375 frames let's go ahead click open and it's reading it does not showing you that is read but it is will read a lot of a lot of frames as we need it click generate the application will generated files inside your output image to images and in a folder when it's generated it will read your first file okay with this directory is empty you can create subfolders like what I did it and copy and paste inside those subfolders if you need it but generally this first need to be clean so it will create your first image and reference that image and you can see right here this is our reference image and our first frame so now it will generate different type of frames okay if you open a folder and look inside you'll notice how the system works we'll create our first frame as we have here and after you'll notice also it will create this kind of reference frame so you have your first frame second and this is called by blending working with consistency by reference and you can see it will create another frames and have it more consistent result by comparing those frames and be sure the all look very identical process is usually take very very long time I'm using on a 3090 video card and it's took for 375 images frames it's took about four or five hours so I highly recommend run it before you go maybe sleep or somewhere else you're going to do or you can run something other because it is will take a long time after render completed you should have a folder with a all frames and sequence kind of done so you can go ahead and preview all of those frames now we need to put them together for this one we're again going to use it same shutter encoder only now we're going to combine images so I'm going to select all of the frames that I'm going to use select them here and drag and drop inside okay actually let's go and not remove my movie that I had before right here all of them set let's go to be sure we select all of them inside now as functionality I'm going switch from creating you remember um images so we want to actually create now animations and I won't go with output codex h.264 it will create MP4 right here same you can say where you want to create it in a path a specified path for this and when we're done let's go ahead just click Start function should process them quite a bit fast and after you will have it file done with your name or you have a frame and you can see right here this is animations put it together so it's got a nice free applications again link for that applications will be down below you can download it for free and you can extract frames or putting together stuff but this is actually only one way one of the biggest things plus of the script because it's acting as a batch file processing it can work almost with any other extensions that we can use it inside the image to image in this case I'll show you how with a rope or face swapping it can work as well so for this we're going to create we'll take it okay and this is image of me so it's what we're going to use it next I'm going to type maybe um and also let's readjust our width and height to match our image same things with almost we done before well let's go ahead and click on generate this should be just generate image without actually references right here I want to be sure it is image of the some Rococo Baroque Queen it's creating for us next right there what we want to do it is be sure the pose is look same we're going to utilize same as before control net very very similar what we've done before and I'm going just copy so we have it there enable let's go Pixel Perfect one of the things about Pixel picture if we enable preprocessors let's enable here if we preview on this you can see how it will start protecting background this actually work very well in this case and beside protecting you can see it is replacing my face very well with a different model so this is what I want to do then we're going to the next control net actually let me verify I have it all disenable yep it's look properly okay let's go ahead next control batch and this one same I am enable and I'm going to use it open pose as well for this just a little bit reduction very very similar to what we've done before and that is to keep consistency okay now this is what actual video of the script coming up that we can also use it like group or face swap so I'm going to expand this you can use it portrait run here we'll go to of course enable default let's go use it GF pen visibility upscale I think everything is right at this point we said and let's go ahead generate with a consistent result and with this face look right here this is exactly face what we have there I mean it's very nice how it's done job because we can even create little bit more smoothness of this everyone can control net as well enable again Pixel Perfect let's go to the model select our temper which is work very well when you do with the portraits and we're also going to control weight let's set about two around this and control net so this way we can create a very nice and smooth okay right here you can see it's make quite a bit better blending so when you like and satisfy with what you see on the pictures we can go start animating so let me go on a script as before and we'll do select script that we're going to use it as well on this script we want to match our the noise string 7.5 we also want to be sure our is selected first generated version rated here enable and as well we want to go to select what frames we want to process an example here we can go select frames Okay click open and now we're ready to go ahead and press as well next going and swipe faces with the model this is in some case you can not necessarily need to use the somebody face let me show you just example we're just going and use a terminator and you know what on the script actually I want to disable because I don't want process I just want preview right now so we'll use it here and we also let's have just rename it and create Terminator if I only can spill properly okay let's go ahead click generate and here we have our Terminator face instead mine so you can see how fun we can create all of these animations and we'll do when we do face swap we can have it nice and you know what interesting things to me look right here this is our original and it's look nice and this is only have the one image to extract and look how clean and nice of course we have it little bit in consist to work but this is example on a face swapping so hopefully you found this video interesting to you you learn new techniques if you do please give us thumbs up like it subscribe share this video help my channel grow and this is the best support you can provide I appreciate all your help and have a great time

Channel: Vladimir Chopine [GeekatPlay]

Views: 27,098

Keywords: Geekatplay Studio, Vladimir Chopine, Digital art, AI art, MidJourney, Stable Diffusion, Dreambooth, Dall-e, Free resources, Free learning, Digital art for begginers, Free tutorials, artificial intelligence, stable diffusion img2img, stable diffusion animation, stable diffusion video, stable diffusion lora, stable diffusion controlnet, ai art generator free, ai animation video generator, character animation blender

Id: 9um0vDJ7_Q8

Length: 22min 28sec (1348 seconds)

Published: Mon Jul 03 2023