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
FAQ
Comments (15)
那么多“全局连接”节点，应该是有冲突，至少我测试是不工作
Sorry you had trouble - if you have all the correct custom nodes installed and updated then there should be no connectivity issues. If you post some of the errors then I might be able to advise more.
Hey hey , thanks for the workflow its working great , one quick note I downloaded the workflow but there was some nodes that were not connected, the vae decode and negative, I then downloaded the video you created and added it to the workspace to have it properly connected so you might want to update the download link with the proper fully connected workflow, but except for that its fast amd great quality outcome , thanks !
Thanks for the message. I will check the uploaded version!
@TheFatController awesome and also you should definitly add a version with a second video like the workflow from your video , I just tried it and it works great , I reduce to 24fps and got a 20 second video in 11mns . I m going to try to see if I can add more and really extend the video but it s a great feature
Strange - I just downloaded the file from the website and loaded it up and for me they are connected fine! I wonder if this is just a weird bug from when you first loaded it... I think the "Everything everywhere" custom nodes are fantastic but a little buggy when someone downloads the workflow fresh.
At least your comment might help anyone else with the same issue!
@lopezd20 Ah yes - I hadn't tidied things up when I generated the video that has been posted. I was testing grabbing the last frame and then having it generate a second video. It does work, but it's not as good as I had hoped and makes everything a lot more complicated so I decided not to include it in this version. Perhaps I will do one in future.
If you upload anything you created, please add this workflow as a resource so that it shows up on this page and I can see it, I love to see what others make using my resources :)
@TheFatController oh ok for me I just unmuted the second video group and added a second prompt and it worked great , wasn t too complicated. I m thinking of trying out creating switch so you generate a video save the last frame and send it to the second , if you re happy with the video deactivate the first and start working on the second and so on. Always had an issue with the extend workflows as you have to enter multiple prompt from the start and if one of the video comes badly all the one that follow will be the same , so doing the extend step by step checking all the video to make sure they are good would be awesome. Never really got into properly understanding the workflows but I will give it a shot as yours has a lot of potential ! 🥳
im new to this but i tried to get it working.
i drag and dropped the .json file i downloaded here to my comfyUI. than i had do download a lot of stuff which i did.
when i try to create a video I get this error:
Prompt outputs failed validation: Required input is missing: images VHS_VideoCombine: - Required input is missing: images
Seems like some notes are not connected like in this comment?
I also tried to download the video here and drag and drop it to comfyui but there i get this error:
Unable to find workflow in Test_00005.mp4
what can i do about it?
@asatru081 at the time I fixed it by downloading this video https://civitai.com/images/112642759 and adding it to comfyui , the workflow i got from the video was properly connected. So download the video, drag and drop it to comfyui and you should have a properly connected workflow
Thank you for this workflow. I tip 500buzz for this. It comes just at the right time for me because I was struggling with my previous basic I2V workflows samplers for some reason and alternate samplers did not fix it. Every workflow should be like this one : simple, with notes, and direct download URLs. Nice usage of 3x samplers. I tweaked the workflow to suit my needs as follow: 1) I added SageAttention nodes immediately following the Load Diffusion Model nodes and this increased the overall speed of the workflow by 30%-50% 2) Using Pytorch fp16accumulation nodes immediately before the KSampler nodes seems to help with heavy loads 3) I am using alternative checkpoints which already have lightning, step distill & other accelerators baked-in so I removed the lightning loras because they were conflicting with that and causing artefacts. 4) I replaced the Latent Video node by the very similar WanFirstLastFrametoVideo node and its related Load CLIP Vision and CLIP vision encode nodes, to allow both a first and a last frame to better guide the video, and I am having acceptable results considering I am pushing it to the limit of my 5070 12Gb : FF&LF 640*896*141frames, 12 steps + 3x upscale + VFI 3x + PingPong gives a 14 seconds video output of 1728*2496*840frames @60fps , in 16 minutes.
Hi! Thanks for the comment and a lot for the tip. I'm glad that you like the workflow. I'd be really interested to see what you make with it if you upload any of your results to this website. Please consider listing this workflow in the used resources so that it shows up here and I can see it!
I will also consider including some of the additions that you made yourself. I have not had much luck getting SageAttention to work, but it would be nice to speed things up. I am also interested in which checkpoints you are using that have the lightning loras baked, because I am always keen to try new checkpoints if they give good results!
Right now I am working on an updated version that includes 4 steps so you can combine 4 clips together to make one long video with multiple different loras and prompts at each step.
It might be the "Anything Everywhere" node being funky. Apparently if you change your language to English (or away from English) and back again it can fix it?
its perfectly solved! you are genius!