Comfyui AnimateDiff v3 + LCM Video to Video

yo guys welcome we got animate diff updated to version 3 more stable and consistent compared to V2 that's awesome so today I'm going to show you video to video conversion with animate diff V3 and share some tips and tricks along the way I know this workflow might seem a bit complex at first but don't worry it's not that hard once we go through it step by step I'll try to explain everything in a simple and clear way so you can follow along and get the hang of it let's get started it is basically composed into two parts the top one is frame pictures preparation while the bottom is animation generation first let's look at this folder text operation I use it to Simply slice and dice texts to provide paths for various image and video nodes put your output folder path here and subfolder names here for different control net outputs split by comma that's all later the these texts can be easily selected as inputs it's a simple way to create and manage the project folder a LIF saer okay let's use these nodes to generate the pre-processed picture for each frame we only need open pose for this example all these pose pictures can be reused for other styles then deactivate the top group let's move on to the animate generation in this part we will use control net an IP adapter to fine-tune the outcome multiple control net can be utilized but let's deactivate this one for our example as well as the output pictures only video is good for now for the input part you can select the image number to be processed I suggested to use small number to try it first like 16 is specified for the context length lower number may cause weird noise then the animate diff V3 here followed by the the LCM Laura the new domain adapter is also added here those pose pictures generated will be imported into this node an IP adapter here can enhance the colors contrast lighting and other features of the image to make it look more appealing there are two pipes used here one for K sampler and the other for face detailer here is another text splitter for the output subfolder then we import one picture to set the size automatically others are General flow another thing I want to mention is the color match node it lets you adjust the colors of your animation to recover or match a certain mood or theme as we did in Magic anime for example we want to create a sunny and beachy Vibe here you can use a picture that has those colors and apply them to the animation you don't need to use the IP adapter like I did here you can use any image you want as a reference and last is the face detailer for animate diff you can check previous video for more info that's the basic process of how it works ready to give it a go and check the results Doo used a animated model now let's change to a realistic model wow this model also looks good especially recovered the realistic face you may try some pictures for the color match node to get good color okay that's all for today hope you enjoy it thanks for watching and have fun

Channel: JanRT

Views: 6,633

Keywords: comfyui, AnimateDiffv3, LCM, controlnet, ipadapter, color match, stable diffusion, consistent, animation, open pose, face detailer, folder name parser, workflow

Id: XQLtI0KT2G0

Length: 3min 50sec (230 seconds)

Published: Sat Dec 23 2023