I just uploaded a LONG video version of this model here: https://civarchive.com/models/2203691/wan-22-i2v-workflow-for-long-videos-clean-upscaling
I learned WAN 2.1 and published a simple 60 FPS i2v workflow for it that I found useful and neat. Now I have finally got round to learning how to get it working with WAN 2.2 and I present that here! I ended up creating this tidy, short, and relatively simple workflow that allows you to perform i2v generation and then interpolate up to higher frame rates.
Please tag this workflow in any videos you upload here. I don't care about attribution, I'm just really interested to see what others make using the tool!
It was useful to me, and it's what I've used to generate all of my videos so far, so I thought I'd share it.
Please note: This workflow uses some custom nodes that are not natively installed with ComfyUI. I highly recommend you install ComfyUI Manager before you load this workflow. If you do that, any missing nodes will have an "Install missing nodes" button that will help you get set up much more quickly and easily!
If you are having issues with the "Anything Everywhere" node connectivity then please try the following fix: Change your default language to/from English and then back again. This has helped others!
This was generated with ComfyUI version 0.3.76 and the latest builds of all nodes as of 4th December 2025!
Happy to have feedback!
Description
Added GGUF load capability, included First/Last frame generation option, improved the upscaling, and enabled a toggle for 2 or 3 KSamplers. Also tweaked overall usability!
FAQ
Comments (17)
Im getting the error ModelSamplingSD3
'NoneType' object has no attribute 'clone' this is at "step 1 shift" in the work flow.
Hello. I googled the error and there was a reddit thread with this comment:
"You cannot apply CLIP Set Last Layer to t5 text encoder.
While it can be applied up to SDXL, it's fair to say that it's almost useless unless you're using SD1.5.
(In SDXL, it only creates images differently and cannot provide the same effect as in SD1.5.)"
Are you using the text encoder linked in the workflow or a different one?
@TheFatController yes i'm using the text encoder that you recommended. umt5_xxl_fp8_e4m3fn_scaled. "Attribute Error 8 modelsampling SD3". Am I using the wrong text encoder?
@johnL69 No, that looks like the right one. Weird... I will think more and let you know if I come across anything!
One thought - what version of ComfyUI are you using? Googling that "Attribute...." error suggests it might be a mismatch of the nodes and the ComfyUI version
@TheFatController v0.5.1 is my current comfy version.
@johnL69 Ah, I'm using 0.3.76 - I wonder if that's the issue. Any way for you to try downgrading? And/or have you tried updating all the custom nodes with ComfyUI Manager?
@TheFatController I have updated all the nodes through manager. I starting to suspect my 5090 might not be compatible.
It loads. Load clip ----> Latent Video 1 ----> Step 1 shift ----> ModelSamplingSD3
'NoneType' object has no attribute 'clone'
@johnL69 Sorry! I'm not much more help than Google I'm afraid. I sometimes feel like I have no idea what I'm doing and am glad it just works.
I haven't upgraded to ComfyUI 0.5 yet, so that might just break mine too!
@TheFatController i downgrade my comfyui to your build 0.3.76 but i can't get custom nodes to work it with it because they are outdated. I'm going to try to manually find them lol. I really like WAN 2.2 i2v workflow for LONG videos and would be willing to pay for current comfyui 0.5.1 build if your interested.
@johnL69 I'd love to say I could help, but my best guess is that with 0.5.1 some of the custom nodes haven't updated to be compatible yet, which would just break everything here too!
If and when I do a ComfyUI update I will put it here though.
@TheFatController Thank you very much for your quick response for my issues. i look forward to your workflows
isnt the int4 versions not compatable with 50 series gpu?
I'm afraid I have no idea! I run a 4080 and I'm not sure what int4 means :)
Your workflow was great, but it's become completely unusable after the last update. I hope you'll update it soon...