ComfyUI AI: Inject reality into your animations

hello and welcome to my new video through which Charlotte's friendly AI voice will accompany you my original idea for the video was to celebrate my 1,000 subscribers by showing you what my style of prompting is and to put my gathered prompt lists into a workflow that you could have downloaded from my website afterwards but then Mato from the latent Vision Channel published a new node that allows you to influence the individual blocks of image generation via prompt injection this had been on my wish list for a while however after a short test with various models I realized that this would require more time so it came in handy that I bumped into the webcam capture node while checking out the latest comfy updates and immediately integrated it into my IP adapter weight scheduling and animate diff evolved workflow but enough talk let's take a look at what this integration looks like first we need the basic workflow for the adapters and animate diff evolved again I'm using the workflow that I explained in more detail in the video before last so you can look it up again if you have any questions the nodes that are new here are the webcam capture node and an image sender node as well as two image receiver nodes by assigning IDs in the image sender node we can send as many camera images as we want to the IP adapters via the image receiver in the variant I am showing here I have added a load image node but it works just as well with only camera images almost there but since the webcam capture node is a bit stubborn let's take a look at a few useful features of the ad group function this this not only ensures that there is a nice background and that the nodes are kept together in groups it also helps us to switch the individual groups on and off effortlessly via the RG3 comfy settings all we need to do is Tick the appropriate boxes in the settings and then click on Save to be able to use the webcam we have to give the browser permission to do so however it is certainly helpful to pay attention to what the camera is focused on because of privacy and so on as we first need the camera images we can use the group mute function right here now simply click on Q prompt and the image receivers will receive the desired images as previously mentioned we can use the link IDs to determine which receiver should receive which image and that's the digital connection between the real world and AI everything you can capture with your camera can now flow into the video generation via the IP adapters and animate diff evolved one of the great features of the IP adapters is that they are excellent at handling photos taken by a camera whether it's two stones in a picture or a container filled with water and a snow globe in it a few drops of food coloring or various camera filters the IP adapters bring it to the sampler I have have added a control net to the workflow so that the continuity and stability of longer videos that is many frames does not suffer even with many input images all you need to do is download the appropriate control GIF checkpoint via the comfor UI manager in the uncut version of this video you can watch all my recorded and admittedly sometimes somewhat crazy experiments if you like that's it for this video for now if you found it interesting and or helpful I would appreciate a like and a subscribe thanks for watching and as always have a great day

Channel: Show, don't tell!

Views: 1,105

Keywords:

Id: Zyao2Oq_aSs

Length: 7min 56sec (476 seconds)

Published: Thu Jun 13 2024