Video Object removal with AI - ComfyUI TP OmnimatteZero - AI Clean-ups

So let's clean up with AI and let's use confi. So I just released something called TP ominate zero which is based on a model module called ominate zero fast training ominate with pre-training video diffuse model. So basically they are using a pre-train module to fast training free which is means that we're not going to need to to create some loras to do some training. It's working on inference time pace. So it's just should be just running it and got the the video object removal or clean up for us. Right. So I made a confi version so it's easy for us. as you can see. And then you can just take the workflow here. If you wanted to do the if you take the workflow, you should have something like this, which is exactly what I have here. Uh so this is what came with the uh with the example footage with some masking. And then you got the cleanup plates in there. So I noticed something. Uh if you use a mask without expanding, you might get some artifacts in there. You see it? It's because the audio sometimes sees between the masking the edges and try to make sense of what was happening there. So basically this is in painting the mask region based on the rest of the footage. Right? So one way to fix that is just expand it. So like some pixel spread or something. So it's much expanded now. And then you just processing and then we should be seeing now it's like cleaning up all these artifacts for us. So let's see it. So it's running here is processing the 30 steps. So it steps how many noise and then also uh this is running the native weight the n the native module inside confi. So it means that it's not quanti. So it's using lots of VRAM and because of that like we need like a heavy graphics card of course. This one is using right now 95 gigs. This is because this is a longer shot 240 frame shots you see here. And also first time you you run this conf workflow it will download automatic.