NEW A.I. Animation Technique! AnimateDiff + Automatic1111 - Full Tutorial

foreign today I'm going to show you how I was able to make this animation [Music] using the automatic 1111 stable diffusion interface with the animate diff extension all the images that make up this animation were generated using prompts generated by the Tyrant prompt generator if you're interested in getting your hands on this click the link in the description below to join the Tyrant Empire's private Community now with that said let's Jump Right In all right so we are ready to go now the first step of the animation process for this is to find your inspiration now for my inspiration I found this quote by Jen sincero that I am going to use for my narration now you can use a story you can use a song and you can even use a quote but if you're going to use a story or a quote personally I lean more towards quotes because they're usually shorter than stories obviously so to generate my audio from my quote I am going to use 11 Labs if you've never used 11 Labs before it is wild to say the least I will leave a link to 11 Labs down in the description or you can just go to elevenlabs.io 11 Labs is a text to speech generator that has literally hundreds if not thousands possibly more of different voices you can use you can just find a voice that suits the mood that you're looking for I have already found mine and I actually have already generated my voice trying to things step out of your comfort zone take risks do things in ways you've never done them before ask for help surround yourself with self-actualized people become obsessed with the fact that you have one go round on this planet as the you that is you and realize how precious and important it is not to squander that so once you have that generated now you can base your story that you're going to tell via animation off of this audio next we are going to Envision a rough idea of what we want the animation to look like and basically just get a mood for it so for this step you just want to visualize a story that goes along with your audio now for me try new things immediately I just think of like a man working on something basic but it you can you can visualize that so try new things boom that's the first scene and then step out of your comfort zone boom Another another scene take risks boom Another scene so you can kind of see how you can just break this down into individual scenes that you can visualize in your mind so now once you have an idea of what you want you can just literally go into stable diffusion and start generating images based off of what you visualize so as I mentioned a man creating something and boom I have an image to base my first scene off of I do this for every single scene that you have created depending on the specs of your computer it would be wise to keep the image sizes relatively small so for example all the images that I'm generating are 512 by 512 I'm using a 1.5 stable diffusion model because I only have 12 gigabytes of vram depending on how you're sitting spec wise you might have a little bit more room to kind of play around with things but for the sake of Simplicity I recommend staying Within These parameters now becomes the fun part so now that we have our images generated you can literally send them to text to image control net and now we have our first image in the control net here so it's pretty simple just click on tile Pixel Perfect and usually it will default the preprocessor and the model to what you need which is tile resample and then this control model here which is tile and now once you have this in the control net just go up to animate diff enable it make sure it's turned on and then we're going to at least I am going to stick with 16 frames and then 8 frames per second simple math that's going to be a two second long GIF we're going to need the same prompt that we had to generate our image which if you have the image browser extension I highly recommend getting it if you don't you can literally copy and paste the same prompt that you used to generate the image back into the prompt to regenerate and since I'm using the dream paper model bedroom and fast Magna V2 are perfect textual inversions to to use with this model so I'm going to activate that I used sde Keras to generate this and so now that we have all the parameters set just go ahead and generate and you can see that got us this fancy little animation here 16 frames 8 frames per second so the next step is obviously doing this with every single image that you have generated all right so now that you have all your animations generated as I have mine some of them I wanted to add on to I'll I'll show you real quick how I did that so you can see I have this and then it just plays on and continues so I have two scenes that I doubled the length of just because it I wanted to and so to do that you're going to need to take the last frame of that animation you generated then throw that into control net and regenerate an animation again best way to do that is go to image browser find the image so this is the image and then you can see the animation frame by frame and then boom it stops right here this is the final frame and then you can even click into the directory here and you can see all the file names to make it a little easier so right about so here's the animation so you want to follow the numbers after The Hyphen so you can see that this sequence is completely different than this this is the actual image that we generated then this is the animation we generated based off of this image because you can see that these two numbers are not the same whereas these numbers right here the 27723 these are all the same so you can see that these numbers are in the same sequence because you can see 27723 1508 then 277 a bunch of numbers and then 1509 and then 1510 15 11 15 12. so the last four numbers you can see that it's actually in a sequence so we that that way we know that this is part of the same animation whereas this image here is the Standalone image we use to generate this animation so make sure that you pick the last frame in the sequence so boom we can see that this is not the same sequence of this so this is the last image of the first animation that we used so if you want to extend this animation as I've done I took this image through into control net and regenerated to get the second animation so now once I composite these I put the first animation and then immediately the second animation so it's a seamless transition from one animation to the next the reason I go over that is because of the next step so you can see in my animation how I have everything transitioning from one scene to another they're just kind of morphing into each other so we're going to use that same technique I use to extend these clips but we are going create what I call transitioning Clips to transition from one scene to the next now to do that we take the ending frame from the first scene and blend that into the first frame of the second scene so to do that we are going to have to introduce a second control net so enable another control net now if you don't have the option for multiple control Nets just go into your settings here go down to control net and then you'll see this settings panel here you can see this multi control net set your max model amount to whatever it is you want I have mine set to three I don't know the default is nowadays it might be three it might not but if you don't have multiple control Nets this is how you add or remove more control Nets I wonder personally what the max number of control Nets you can have up to 10 bucking control Nets look at that and then make sure you apply the settings and reload your user interface to implement the changes now moving on so in order to create that blending scene as I mentioned we're going to need to take the final frame of the first scene and blend it with the first frame of the second scene and again pretty repetitive just go into image browser or your file directory grab the designated frames plug them into the control Nets and generate so you want the first frame in the lower number control net and then the second frame in the higher number control net because it's going to take what's ever in the lower number control net and that's going to be your starting frame and then the second frame with it which is in the higher number control knit it is going to be what it is Blended into so now go ahead and do that with all of your scenes take scene one blend it with scene two take scene two blend it with scene three scene three to scene four four to five etc etc so now if you did everything correctly you should have your blending scenes and if you merge the correct frames they should blend pretty seamlessly but we won't know that until we throw them into a compositor now it is worth noting that upscaling will be rather important because there's not much use for animation sizes that are 512 by 512 but it's really small so you want to at least upscale them two times you want at least a 10 24 by 1024 I use topaz video AI to upscale they have a bunch of different models that upscale enhance detail and then also interpolate frames to make it buttery smooth so I went from eight frames per second to 60 frames per second so that's why this looks a lot smoother than what it traditionally should but you can see that it transitions very very well alternatively I do know you can use Optical flow in DaVinci Resolve they have a frame interpolation feature in there I have never personally used it so I can't really give you advice yet then I'll also you can batch up scale in automatic 1111 so you can go to the extras tab here go to batch process from directory or just batch process and just drag the file with all of your frames in here and then you can upscale and yeah you just pick an upscaling model and that'll take it from well you would obviously want to set what you wanted to upscale by which would be at least two if you're 5 12 by 5 12 so you can get a 1024 by 1024 or you can do resize it by four and get like a 20 48 by 2048 which is quite large and then from there you can just throw it into either DaVinci Resolve which is a free software or Premiere Pro which is what I personally use for this kind of stuff because it is what I know best [Music] and now just throw in my audio and add subtitles take risks do things in ways you've never done them before now if you've never made subtitles before it's really quite easy you just click on the audio file that you want to transcribe right click and then you can transcribe I've already transcribed this so just retranscribe so after you do that if you don't see a text option here just click these arrows here and then go to text and then you'll see all your text so give it a sec so boom I have all my text here and then we can take this text put it under the video by clicking these closed caption or create caption things here and then you want to click the preferences now this is what I use for all my stuff I do 1.7 second duration and then maximum length in characters at about 15 and then I keep the lines at single so the subtitles aren't stacked on each other that way they kind of keep the viewers attention that way they're continuously changing so the user doesn't really have to scroll to read like they don't have to scroll with their eyes it's just constantly changing like the text is constantly changing I don't know I don't know how many times I have to say that for myself today I've probably said that enough times then we just create captions and then now they are very small these are kind of terrible because they're one super small and they just look very plain so first off we're going to want to change this text because it is boring as hell so I like the Bold font it's very very cliche it's used in a lot of stuff and then I usually keep it around 90 to 100 just so they're nice nice and chunky nice and thick then also turn on the stroke and make it nice and whatever color you want really I usually do black but since this isn't a 9x16 format it's not going to look very shh snazzy so we're just going to do like a blue now that doesn't look bad at all now I also forgot to mention if you're completely new to like Premiere Pro or anything like that and your settings are all kind of whack just go to sequence sequence settings and then this is where you can change your frame size so if you're planning on uploading this to like Tick Tock or whatever as like a reel or a YouTube short you're going to a 9x16 format which is usually 1080 by 1920 where you can see that it's 9x16. or you can just do a one-to-one and then just set the dimensions to that of which your images or animations are set to and now that basically completes it so I didn't add music to this because when I upload this to Instagram or Tick Tock or Facebook or even YouTube shorts I want to use trending audio because especially on Instagram no one sees like my reach is so bad on Instagram so if you're watching this and you want to follow me on Instagram link is down below but literally I think I post I think I get like 20 to 100 of my followers that actually see my stuff it is what it is but trending audio you actually get to kind of put it out there in a different way than you would if you just throw you know generic music on here and then post it so if you don't put the music in the composite it just kind of gives you a little bit more room to kind of experiment with different stuff like in an app so that's what I'm doing you can do whatever the you want yeah and that is the process of how I created this fantastic animation and and since I delve into this which has been like three or four days there has already been a ton of other improvements in different areas of AI so I have quite a few ideas that I'm going to already create after I upload this so if you like this kind of stuff and want to support me and my channel definitely subscribe because that is how you do that or if you want help feedback be in a community of like-minded individuals who are trying to better themselves in every aspect or Master the Pinnacle of digital art creation join the Tyrant Empire Discord with again the link is down below and with that said I hope you have a fantastic rest of your day wherever you are peace and love stay safe out there I love you very much later

Channel: Ty The Tyrant

Views: 16,636

Keywords: tyrant, empire, tyrantempire, rampayj, ai, artificialintelligence, art, midjourney, stablediffusion, automatic1111, leonardoai, dalle, chatgpt, fitness, selfimprovement, selfhelp, learning, gaming, apexlegends, leagueoflegends, callofduty, dating, women, redpill, entreprenuer, freelance, freelancing, money, business, memes, streaming, tyrin, music, musicproduction, dj, howto, makemoney, america, usa, lamborghini, ferrari, bugatti, honda, funny, muscles, abs, sixpack, 6pack, diet, food, polymath, books, reading, god

Id: HAAC36X-HEM

Length: 15min 17sec (917 seconds)

Published: Sun Sep 24 2023