Introduction and setup

0:00 You can download all these workflows for free from Discord. Let's start with this Cany version. This is the workflow. It looks complicated, but you'll see it's actually quite simple. I've added everything you need to run the workflow here in this Pixaroma note. If you've used Flux before, you probably already have most of the models. The first model is a Fluxdev model. I'm using the Q8 version and you can download it from here. Then go to your Comp UI folder, find the models folder, open the diffusion models folder, and place your model there.

0:48 If it's too big for your video card, you can click here to download a smaller model like the Q4 version or another option. Next, you need two models for the dual clip loader. Use the GGF versions because they are faster. You can download the T5 model from here or use smaller alternatives from this link. You'll also need the clip L model. I just download both models and load them here. You'll also need a model for the load VAE node.

1:12 You can download it from hugging face. If you get a message saying you're not logged in, it means you need to log in to hugging face first. Click here to go to the model page. Log in and accept the terms. After that, you'll be able to download the VAE model. Again, if you've used Flux before, you probably already have it. Now, the most important part, the control net model. Click here, go to your models folder again, and this time place the model in the control net folder.

1:58 But first, rename it. If you have multiple models, they might have the same name. I renamed mine to Flux Line CN Line Pro line 2. Whatever name you use, make sure you load it with the same name in this node. And the last step, you need the custom nodes. I used only three custom nodes here. Go to manager and then to custom nodes manager. Search for GGF and install that node. Search for AUX and install that one too.

2:22 Then search for K tool and install it as well. After that, restart Comfy UI. If you go to settings and then to K tool, you'll see that by default this Gulu or whatever it's called option is active and it shows a cute dinosaur. If you don't want to see it, you can disable it from there. Okay, now you have everything you need to run it and we can test the workflow. Upload a reference image here. I'll use this building image.

2:47 Then I use the scale image to total pixel node to make sure it's not too big and works well with the flux model. I set it to 1 megapixel here. So when I run the workflow, whatever the