Transfer Motion From Video to AI Character LTX 2.3 | ComfyUI Tutorial

Hey everyone, in this video, we're going to see how to animate a character from an image by transferring motion from another video using LTX 2.3. For this, we're going to need a Laura called IC Laura Union Control. The download link for this Laura is available in the description. If you open the link, you'll see a page like this. This Laura supports multiple control conditions such as Canny, Depth, and Pose, so it can be utilized in various ways to copy composition from another video. To download the Laura, scroll up and click on files and versions. Here is the Laura named LTX 2.322BIC Laura Union Control Ref 0.5. Use the download button to get the file and place it inside the Laura's folder in ComfyUI. I've already done that, so let's switch over to ComfyUI. So, I've already set up a workflow that you can download from the description. This workflow uses several custom node packs including ComfyUI Easy Use, ComfyUI KJ Nodes, ComfyUI ControlNet Aux, ComfyUI LTX Video, and ComfyUI Video Helper Suite. You can install all of these through the ComfyUI Manager. Now, I'll give you a basic overview of the workflow so you can start using it. So, here we have the load image node for loading the image that contains a character. Right now, I will be focusing on only one character, so let's choose another image. Let's choose this photo of an ancient woman. Then we have the load video upload node for loading the video from which we are going to copy the pose. Also, this workflow will make sure the output video orientation matches the video we load. I also realized that if we change the pose of the character in the loaded image to match the pose of the first frame of the video, we will get better results. So, for changing the pose, we will be using QN Image Edit 2511. Here I wrote a simple prompt for transferring the pose from the first frame of the video. For extracting the first frame, we are going to use the Get Image or Mask Range from Batch Node from ComfyUI KJ Nodes. The extracted frame will go into the AIO Aux Preprocessor, which will create a control map and then send that control map to the model. Right now, I am using the Open Pose Preprocessor, but you can also use the DW Preprocessor. So, let's change her pose. I'm going to click on the Save Image Node, then click on this small play button and wait for a while. So, here we go.