⚠️ The I2V Vid extend workflow is highly experimental and requires high end GPUs.
I've created this Wan Video Image2Video workflows based on Kijai's workflow and the Native ComfyUI workflow, added an easy to use interface along with upscaling and frame interpolation.
Key Features:
1st pass - generating vanilla video
2nd pass upscaling using user selected upscaler and frame interpolation to 32FPS
This workflow comes pre loaded with my Wan14B RunPod template:
https://civarchive.com/articles/11960/wan14b-in-1-click-with-workflows-included-runpod-template
Join my Discord server for updates on new LoRAs, tips & tricks, Workflows, RunPod templates and the holy right to be close to an amazing charismatic person like myself:
https://discord.gg/fyha5Pzm
Description
Added 60 FPS and power lora loader
FAQ
Comments (41)
How much GPU RAM is recommended?
24 +
works with 16 too, just a matter of time :)
Damn !! It so fuckin good result !! i tried in RTX 4090 24GB cant handled it .. but running well in RTX A6000 48GB..
is there anyway to increase more than 3 sec?
Set frames to 81 for 5 seconds
@HearmemanAI where is the frame parametr?
ehsanbbaxe Maybe a bit late. But just in case its useful for others: Its in the WanImageToVideo node (Global Settings)
I red that by using SageAttention it will speed up generation speed by up to 30%. But using that with the Wrapper nodes, it takes an infinite amount longer compared to vanilla comfyui components. Am I missing something?
Not sure if this is still useful to you: but I found that when things were take 1+ hour(s) rather than 3-10 minutes (using a 4090) it was because either
a) Either one of the nodes was offloading tasks onto the CPU rather than the GPU (a common tactic to save VRAM). Look out for 'cpu' rather than 'gpu' or 'cuda' (thankfully most 'default' will get set to gpu).
b) Or something was set wrong (TeaCache being set too low is a usual suspect) which triggers a constant feedback 'thrashing' effect. TeaCache in particular is mostly worth hitting 'Bypass' on as its not usually worth the hassle.
If I'm using the tiled VAE decode due to memory errors, what should the settings be? tile size, overlap, etc.
The node in the workflow shows no settings, but can't be connected. I can only see the output once I reload the node, at which point the settings show up. Not sure if these are correct settings or default values?
Upscale factor, INT. MIssing this node and cant find anywhere
Go to your node library and search for "int". Anything basic with a value field and an int output will work, such as utils/primitive/int.
install comfyui-logic node
You need to install the comfyui-logic module via the ComfyUI Manager.
In order to be able to so, you will need to {install-location}\ComfyUI\user\default\ComfyUI-Manager\config.ini and change 'security_level' to 'normal-' e.g. security_level = normal-
So do this with discretion, and consider changing it back afterwards.
@toffee_v That didn't fix it. Exact same error as before.
this is amazing
Awesome! Do we have a chance to get a dedicated version of v2v just for upscaling?
buy topaz video ai , is awsome
I am trying tu run this with lightx2v lora and while succesful in massively reducing generating time, it ignores other lora used with it. Why is this happening?
Sadly does not work for me. The first pass works great and generates a video, but the second pass Ksampler just does nothing for hours, like something crashed in the back.
RTX 5070ti 16GB VRAM, 64GB System Memory.
Even when using with fp8 model, same issue :/
idk if that's your issue, to run wan i need minimum 24gb vram. I'm using a 3090 Ti, and wan works ok on 23/23.5 GB of vram full (q6), but if I switch to q8, or it crash or it needs hours instead of few minutes with q6. What I experienced (hours for a video) is when you are trying to load something that is bigger than your vram. And that's exponential, eg if with q6 who fits in my 24gb vram i need 15 minutes, if i use q8 it may need even 1.5/2 hours, and often stuck in model loading. So yeah, i think for video generation with the minimum is 24gb vram as standard, better if you have 32gb of a 5090, as a consumer level. Otherwise a 48gb for even more quality, eg L40S, but that's quite expensive. Try hunyuan, it can fits smaller vram, or smaller quantized versions, eg q4.
Btw I cannot load wan on my second pc with a 4070Ti with 16gb vram, too little.
its a memory issue, i'm pretty certain 16gb vram is too low. When images/videos generate and it uses up 100% of the memory capacity, it will just get stuck forever and nothing will happen.
@haenlesn937 I have the same graphics card as you. Try this workflow out: https://civitai.com/models/1474890/wan-i2v-with-upscaling-interpolating-and-smoothing It works on mine.
doesn't work... used it on runpod with this template "One Click - ComfyUI Wan2.1 - Wan 2.2 - CUDA 12.8" and it gives an error https://pastebin.com/Gs570z2i
also i get this one https://pastebin.com/xwy4n9At
Runs OOM on a 5090
Getting the same Int error on the upscale factor as many others. Installing the logic node doesn't do anything.
Works brilliantly. Was an INT node that I had to fix, but nothing the manager can't do. I love the added loRa Loader, and the way it can stack between modules. 10/10
what was the Int node ? Manager doesn't seem to find it on my side...
@charlesdelavigere743 Exists
Great workflow. Runs incredibly fast on my rtx 4070. Appreciate the work you put into it, required no modification from my end and ran on the first try
seems like i updated comfyui and now the cg everywhere nodes stopped working. Damn this comfyui running keeps being difficult :(
The extension isn't good at all due to the issue
This is amazing running it on my 5080 and I can get a 4s video in 54s
This is the best set of workflows I found on civitai so far. I love that every workflow is separated by function, unlike many "all in one" flows where it takes forever to understand the logic. Thank you
For a no bullshit model...well, it doesn't work at all. It keeps asking me to connect the positive and negative missing connection nodes
...just connect them lol
Nope, not working.
I'm just learning about comfy and its node noodles, but I see the anything anywhere node is breaking most of it, since it says [depr] I believe its deprecated, and it isn't linking nodes correctly rendering it unable to execute properly, perhaps you can consider updating it, as it isn't being automatically replaced. :)
miss
Set Video Length/Width/Height WanImageToVideo kết nối đầu vào positive và negative