Wan-Video-Extender Node for ComfyUi
I created a node for infinite video generation for wan2.2 that keeps VRAM flat by offloading segments to disk. It enables dynamic storytelling with per-loop switching of Prompts and LoRAs to evolve the scene and style on the fly. It works perfectly with WanRapidAllinOne
please share the workflow, thank you for your hardwork
Neat! Curious how it compares to this one I've pinned for awhile: https://huggingface.co/Phr00t/WAN2.2-14B-Rapid-AllInOne/discussions/100
Neat! Curious how it compares to this one I've pinned for awhile: https://huggingface.co/Phr00t/WAN2.2-14B-Rapid-AllInOne/discussions/100
I wrapped the whole process into one standalone All-in-One-node, memory efficient ,userfriendly , and many more features, I'm currently fixing some bugs
@test53100 it is a custom-node but the workflow is at my github-repo https://github.com/Granddyser/wan-video-extender , I recommend to wait for my next updated version
great job man! waiting for your next updated version and test result.
your node can't be installed by comfyUI desktop
you have to set the security level n your ComfyUi-Manager folder in "/ComfyUI/user/default/ComfyUI-Manager" open config.ini and set security_level = weak
how do you install sageattention
sorry @StevieTee to interrupt your chain-of-thought here, but I managed to install that node even without the lowering of security level. Actually when it asked me to choose the version, when select 'nightly' it gives me that annoying error mesage, but when select 'version 1.0' it installed the node... just to let you know, you know....
how do you install sageattention
(uv) pip install sageattention
The main issue I have , I still haven’t figured out how to make the node versions and last update timestamp visible in the comfy manager
Is it possible to add multiple loras per loop? Right now I am using a "Lora Name" node to get a list of my loras to apply it to the appropriate loop, it is unclear if the box will only take a single lora or if I can separate them with a comma.
Is it possible to add multiple loras per loop? Right now I am using a "Lora Name" node to get a list of my loras to apply it to the appropriate loop, it is unclear if the box will only take a single lora or if I can separate them with a comma.
I initially just passed the LoRA path as a string when I created it, because at that point I hadn’t really looked into how LoRA is actually loaded in the ComfyUI core. My main focus was on getting the iteration part working first. In the future, I’ll add proper LoRA connections directly to the node.
For some reason I can't get this node to work at all. No matter what I pass to it (image, video, reference image, etc) it always generates something random and unrelated to the original input. For example, if I feed it an image then the image pops up as the first frame of the video and then the very next frame is a random, crappy scene completely unrelated to the first frame, and it continues from there.
Also, every subsequent loop does the same thing - it starts again from scratch, ignoring the previous loop it did. So basically every loop is just a brand new, completely unrelated video, including the first loop. Been trying all different combinations of settings and nothing changes.
Using the latest version of comfy. Any ideas?
The high/low noise workflow has issues for me. Im using Fun Vace Module + Wan 2.2 I2V KJ Models with 4step lora. My output is full of noise. Any suggestions ?
For some reason I can't get this node to work at all. No matter what I pass to it (image, video, reference image, etc) it always generates something random and unrelated to the original input. For example, if I feed it an image then the image pops up as the first frame of the video and then the very next frame is a random, crappy scene completely unrelated to the first frame, and it continues from there.
Also, every subsequent loop does the same thing - it starts again from scratch, ignoring the previous loop it did. So basically every loop is just a brand new, completely unrelated video, including the first loop. Been trying all different combinations of settings and nothing changes.
Using the latest version of comfy. Any ideas?
Sorry, seems like my previous message didn’t actually get posted. I accidentally only quoted your comment without my text being sent.
What I wanted to say was the following:
Would you mind sending me either
a) the full terminal output, or
b) the exact workflow you used, including the node settings and connections?
That would help me narrow the issue down properly. Thanks.
The high/low noise workflow has issues for me. Im using Fun Vace Module + Wan 2.2 I2V KJ Models with 4step lora. My output is full of noise. Any suggestions ?
Same four you , I need the workflow your using to identify the issues
Hey, just realized my previous message didn’t go through. I accidentally only quoted your comment without actually posting my text — sorry about that.
What I wanted to ask you was this:
Could you send me your full workflow exactly as you’re using it? Including the node settings, prompts and the connected input nodes. Also, if you have any terminal output, that would help too.
With that info I can try to reproduce the issue on my side and see why it keeps generating unrelated frames.