How to download step by step?
New here to Gitbub and have no idea how to download this program
Can someone please give a easy to follow, step by step break down of how to download Wan2.1-I2v-14B-720P
Hey man... or woman, umm well that depends how you want to go about it. I think the easiest way which doesn't involve any terminals and code would be to download an application called ComfyUI Desktop. I am currently using that one so I know it works, you see your other options are actually doing a little bit of coding and using the gradio UI, or you could use the version of ComfyUI which requires you to setup Docker, build it in a container etc. May I ask what machine you're planning to run this on? I have an NVIDIA 4080 super and still needed to download the Q4_K_M (quantized/compressed) version of the model you are planning to run, and it still takes idk long enough to go have a coffee or two. Anyway regardless of what model you pick the first step is to go to www.comfy.org then download and install the correct version for your machine. The simple one click installers require you to have nvidia gpu on windows or apple silicone on mac. Once that installs you should follow this guide, and download the comfyui models, it will be much easier than other options (probably). Let me know if you get stuck and I'll help you out. The best bit about comfyui is that even though it looks complicated you can frag an image of someone elses workflow and drop it into your comfyui, and as long as you have things saved in the correct folders, and the correct nodes available it should work: https://comfyanonymous.github.io/ComfyUI_examples/wan/ Make sure you place the files in the folders like it says, these should just be in C:\Users\smith\OneDrive\Documents\ComfyUI then click on models, and download the required files for the model you want to run into the folders stated. If a folder is missing, just create it inside of the models directory, name it, and you should be good. Now for this stuff to become available you should close comfyui and start it back up again. Once you're at this point you can drag one of those images in the link I sent you and drop it into comfyui, and a pre-made workflow will appear. I'll wait for you to get back to me, but you're going to need a really decent GPU (or M whatever with plenty of unified memory) to run that model you chose.
https://huggingface.co./city96/Wan2.1-T2V-14B-gguf/tree/main
This is where you get the quantized models, pick one that will fit on your GPU or your unified memory if you use a mac I guess, imo anything between q4 and up will be good quality, don't bother with q2 or something like that, just get the regular small model in that case. Only difference here is that you have to save these ones in the Unet folder (which is also in the models directory), and you have to go into manager in comfyui, then custom nodes manager, and install ComfyUI-GGUF node pack by city96 (he's actually the same guy that has the quantized models on hugging face). You just have to swap out the node which contains your Wan2.1-I2v-14B with a new one from that pack you just got called Unet Loader (GGUF). Right click on the one you want to remove, then click remove. After resetting comfyui you can find the replacement in your node library (far left of the page, looks like a book symbol). You'll see it in my screenshot, it's the first one under bootleg dropdown menu. Linking up the noodle thing is just a case of clicking a dot, dragging the cursor until you're over the one you need to connect to on the other node, the letting go.