Deforum for AUTOMATIC1111 - Stable Diffusion Tutorial - AI Animation Part 1

Video Statistics and Information

Video
Captions Word Cloud
Reddit Comments
Captions
foreign [Music] welcome to this deep dive into the deforum automatic 1111 extension you might have been curious about creating these flashy videos but the deformed column notebook might be a bit intimidating without a proper interface luckily there's an extension for stable diffusion's most popular interface automatic 11-11 in this video we will cover all the relevant functions of this script and show exactly how the parameters influence your animation frame by frame as of now the version of deform we are looking at here is 0.5 so maybe if you're watching this there have been more features added nevertheless the core features should still be relevant for upcoming versions one last thing I am far from mastering this software and have a lot to learn but I will try to visualize all the information I found and Link everything relevant in the description let's Dive In [Music] first things first if you don't see a deform tab in your automatic 1111 interface means it's not installed yet before you install the form make sure to update Auto 1111 with Git pull since the recent update has added the easy extension installer just navigate to the extensions Tab and then the available tab there you should see all available extensions you can install or have installed already click install undeform and it should download the files to your extension subfolder the last thing to do is to restart radio and you should see the deform tab added to your interface check the deform web UI GitHub page for dependencies you need to install for 3D mode on Windows you will also need to install ffmpeg and Link it to path if you want to export your animation as an MP4 video file you will find links to both dependencies in the description if you want to keep track of the progress of the animation you should activate the image creation progress function just enter the settings menu navigate to the user interface section and set the slider to the step value you want to get an updated image of [Music] just a final troubleshooting info every time I started up Auto 11 11 for a new session I noticed that nothing was happening when I tried to generate an animation into forum the solution was to deactivate deform restart radio reactivate deform and restart radio again after checking if everything is working you're ready to animate [Music] [Applause] most of the Run settings should be familiar to you if you have been already using stable diffusion or automatic 11 11. they're the same ones you will find in the main text to image tab like size seed sampler and steps so let's focus on the settings that are relevant for deformed specifically at the very top you will see that import settings from file section where you can link a text file with all the settings for your animation to use instead of the parameter values the settings in the file look something like this when the file will be saved every time you generate an animation if the checkbox saves settings a bit further below is activated right below is the save samples checkbox that activates saving all frames of the animation in the output folder the settings file comes very handy for sharing your inputs with others or to look up the values of an animation you created some time ago it can also make sense to set up your animation in a text file since you don't have to switch between tabs and have a quick Overlook at all the data on the right side of the interface you have also the possibility to load the content of a settings file into the input fields at the bottom you will find the batch settings for saving and naming the files the batch name value will be the name of the subfolder for your files and located in the outputs image to image folder and at file name format you can Define the Syntax for your generated files the next one is very important seed Behavior this setting will influence how the seed value is behaving with every processing frame of your animation [Music] Ida stands for incremental change and will progressively change your C by one for the next frame it will start from either a random seed or a seed you set in the seed input field fixed will use the same seed for every frame of your animation the setting is giving you worse results in animation mode leading to overblooming the images but it's perfect for video init mode to give it a more consistent animation random will use a different seed for every generated frame throughout the animation and can sometimes lead to more chaotic results schedule will activate the function to control which seed to use at a specific frame in the animation we will take a look at the input field for that later in the video before we proceed to the next and most interesting section of keyframing camera movements we need to take a look at the keyframing syntax first you will notice that all values in the parameter Fields look something like this the first number indicates the frame of the animation and the second number indicates the value of the parameter having only one value starting at keyframe 0 will lead to using that value for every frame throughout the animation so if we put the value for rotation Z to zero and three the camera will turn counterclockwise by three degrees per frame for the entire length of the animation if we want to vary the movement we can enter a second value after a comma let's say 60 and 1. now the camera will start rotating counterclockwise by three degrees per frame and gradually slow down to one degree per frame until frame 60. from this point on it will keep the value of one degree until the end of the animation if we want a more abrupt change we can enter 59 and 3 between these parameters in that case the camera will move by three degrees per frame until frame 59 and start rotating by one degree from frame 60. just remember that the first keyframe has to be zero and sections must be separated by commas foreign [Music] keyframe values as we like and even automate the movement with math equations which we will cover later in the video let's move on [Music] in this section we will take a look at how different parameters are influencing the generation of new frames the animation load drop down menu lets you choose between 2D 3D and video input depending on what mode you choose some input fields are being ignored when running the animation generation we're going to look at the relevant fields for each animation mode in another part the max frame's input field determines how long your animation will be to define the length in seconds you first need to know what frame rate in frames per second you want the final video to have the default value in the video output section is set to 12 which is fine if you are planning to make it smoother with interpolation later so for a 10 second video we will need to generate 120 frames in the Border menu we can choose between replicate and wrap this parameter controls handling method of pixels to be generated when the image is smaller than the frame rap pulls pixels from the opposite edge of the image [Applause] while replicate repeats the edge of the pixels and extends them animations with quick motion May create lines where this border function was attempting to populate pixels into the empty space try out both modes and see which one gives you the best results now let's take a look at the general settings that will be applied to any animation mode you choose noise schedule determines the amount of grain to add per frame the higher the value of the grainier your picture will be which leads to more diversity in your images [Music] strength schedule sets the amount of presence of the previous frame to influence the next one a high value will produce more coherence in the image generation whereas a low value leads to more variety in every image this value also controls the amount of steps used after the first frame and you can calculate the steps value using this formula the contrast schedule adjusts the overall contrast per frame [Music] in fact we'll quickly overbloom or darken your image but can be used for fade outs or transitions [Music] foreign [Music] scale schedule lets you control the CFG value on any frame in the animation higher values keep your image closer to your prompt but can also lead to more contrast in the image [Music] foreign with the seed schedule parameter you can determine which seed to use in any frame of the animation as we saw a bit earlier the seed behavior in the Run settings needs to be set to schedule the fusion Cadence controls the frequency of frames to be affected by diffusion setting the value to let's say 4 means that every fourth frame will be generated with the diffusion values we just looked at the rest of the frames will be generated following the 2D or 3D settings while being more coherent to the original frame this can lead to more consistent animations but also lose diversity and result in various artifacts when using a high value skipping the frames for diffusion also means faster animation processing and can be used to quickly check the camera movement settings before applying them for the final animation output the color coherence will attempt to sample the overall pixel color information and Trend those values analyzed in frame 0 to be applied to Future frames [Music] as you can see any of these settings will influence your animation results even with small value changes as every animation is different it's best to just play around with the settings and adjust if you don't like the results [Applause] [Music] in the next video we will be covering the camera parameters for 2D and 3D mode in detail if you are curious about more video examples please visit the deform Discord where many creators share their insights and settings you can find the link in the description and if you found this video helpful please consider leaving a like comment or subscribe to this channel thank you for watching and see you in the next video
Info
Channel: Fictitiousness
Views: 113,866
Rating: undefined out of 5
Keywords: stable diffusion, stable diffusion tutorial, stable diffusion ai, stable diffusion ai art, ai generated art, ai art generator, stable diffusion art, stable diffusion installation, stable diffusion Gui, stable diffusion online free, stable diffusion 1.5 install, stable diffusion inpainting model, stable diffusion download, stable diffusion automatic 1111, stable diffusion de forum, stable diffusion de forum animation, stable diffusion de forum local, deform animation
Id: R52hxnpNews
Channel Id: undefined
Length: 11min 45sec (705 seconds)
Published: Sun Nov 13 2022
Related Videos
Note
Please note that this website is currently a work in progress! Lots of interesting data and statistics to come.