Everything shoved into single node.
Pros:
Independent of any workflow, can be used anywhere.
Versatile. Can be chained for whatever effect: T2V+I2V extensions (unlimited)+FLF2V loop+V2V Upscale+Face Inpaint, whatever else, in a single workflow.
Cons:
Newbie unfriendly - you have to know what and how Wan does to better understand how to use it.
Lora selection - have to be performed inside the subgraph. Easy to forget and miss when they were used, but also easy to copy to the next part.
Easy to mess up inputs and parameters
RAM bloat - using multiple features with different checkpoint means comfyUI will not unload them from RAM, eventually spilling to swap file. For me, even 128GB is not enough sometimes.
Workflow json file bloat - inner wflow is massive with all the branching.
To start using, select your models (ckpts and lightning loras) inside subgraph (relevant nodes highlighted by red in top left corner), install triton and sageattention, or bypass/remove torch.compile and sageattention nodes (be a quitter and lose ~40% speed boost). Publish subgraphs (book icon) to use anywhere.
When using merges with lightning loras included, bypass or remove lightning lora loaders inside subgraph.
There's a pdf manual in the archive to help navigate subgraph's parameters. Also contains short instruction about triton and sageattention installation.
Changelog
v4 2.11.2025
Combined all Wan 2.2 and 2.1 basic features into single subgaphs.
Gathered most of VACE features into sigle subgraphs.
v3.1 17.08.2025
Added:
FunControl 2.2 - inner wflow and subgraph
Changed:
How loras added; initially i opted to using lora stacks, as only 2.1 loras were availible, to select lora once and apply to both ckpts, but now it's no longer relevant, so back to PowerLoraLoader
FunInpaint was very disappointing, so i won't be adding it. I2V is same in function and has no risk of conflicts with loras.
v3.0 01.08.2025
Added:
Wan2.2 model profiles and new nodes utilizing it's MoE capabilities
Changed:
Group nodes replaced with subgraphs
Removed:
1.3b control loras
v2.8 14.07.2025
Added:
VACE seamless extension - wflow and group node
VACE seamless loop - wflow and node
Removed:
HunYuan
LTX
v2.7 17.05.2025
WanWrapper workflow updated
VACE 14b and 1.3b
Phantom
FantasyTalking
CausVid (3step) lora
Spline dot control example
v2.6 24.04.2025
Wrapper IV2V node - crank up blocks to swap and upscale to XL resolution
FLF2V - official start-to-end frame model that rocks
Skyreels V2 DF - video extention model that takes last frames of provided video and seamlessly extends it - pretty much FramePack, but Wan
v2.5 7.04.2025
Wan wrapper consolidated - separate workflows are likely better, keeping it as proof of concept
v2.1. 4.04.2025
InterpV2V nodes added.
v2. 3.04.2025
HunYuan - standard tools: T2V, V2V, I2V (including LeapFusion), IV2V (performs poorly)
introduction of Group Nodes
v1. 31.03.2025
initial release
Description
WanWrapper consolidated workflow
FAQ
Comments (6)
You're an asset to the community - thanks for sharing your knowledge and these workflows/LoRas.
Thank you for your praise!
@cgibson To your question: 'WanVideo VACE Start To End Frame' node has start_frame and end_frame as optional - is it possible to only provide end frame?
Have you looked at Kijai's example (Top group from this file) "wanvideo_1_3B_VACE_examples_01.json" --> https://github.com/kijai/ComfyUI-WanVideoWrapper/tree/main/example_workflows
It uses start and end image ref. I've used this setup for looping animation and it works great with VACE. My only problem is I can't make WAN Loras work with the VACE model. I get errors in the log (like the one below) and the lora does not seem to have any effect. Anyone know if this is just a limitation of the VACE model?
ERROR lora diffusion_model.blocks.0.cross_attn.k.weight shape '[1536, 1536]' is invalid for input of size 26214400
an hour prior to your comment _02 version of this workflow was uploaded to the repo, and it features 'WanVideo VACE Model Select' node that allows usage of VACE alongside regular model.
@cgibson Thanks for that hint - I've tried again with that node and using a "normal 1.3B model" which works fine, but the Loras still don't work and I'm getting the same errors. I guess we'll have to wait for the 14B VACE models for Loras to work?
wait, did you try to use 14b loras on 1.3b? if so, then - yes, of course we'll have to wait) lora should be used on model it was trained for