✨ WAN2.2 — Image to video with flux kontext — Simple Workflow
A clean, all-in-one WAN image-to-video workflow built entirely with the UmeAiRT Toolkit for ComfyUI.
It allows you to generate an initial image using FLUX kontext, and then create a video from that image.
For example, you can generate an image from a person in a photo and then animate it.
Only 12 nodes. No spaghetti wires. Just load your model, write your prompt, and hit generate.
⚠️ IMPORTANT — Nodes 2.0 Required
This workflow is built for the Nodes 2.0 (Vue) interface of ComfyUI. If you don't enable it, the workflow may have display problems.
How to activate Nodes 2.0:
Open ComfyUI
Go to Settings (⚙️ icon, bottom-left)
Find "Use Nodes V2 (Vue)" and toggle it ON
Refresh the page
Load the workflow
🎯 Features
Text-to-Image generation
Automatic download of models in auto version
Built-in SeedVR2 upscaler — high-quality tiled upscaling (toggleable on/off) Slower than a classic upscaler, but significantly better quality
Full metadata embedding — your images are saved with all generation parameters, ready for online publishing and remixing
6 LoRA slots — with individual on/off toggles and strength control and you can connect as many other lora modules to each other for as many LoRA as you want.
📦 Custom Node Required
Only one custom node to install:
Install via ComfyUI Manager (search "UmeAiRT") or use the UmeAiRT Auto-Installer.
The Toolkit packages everything internally — upscaler, face detailer, metadata saver. No other custom nodes needed.
📂 Files you need (in manual version)
For base version
I2V Model : wan2.2_i2v_high_noise_14B_fp8_scaled.safetensors and wan2.2_i2v_low_noise_14B_fp8_scaled.safetensors
In models/diffusion_models
CLIP: umt5_xxl_fp8_e4m3fn_scaled.safetensors
in models/clip
FLUX_Model : flux1-kontext-dev-fp8-e4m3fn.safetensors
in ComfyUI\models\diffusion_models
FLUX_CLIP : t5xxl_fp8_e4m3fn_scaled.safetensors and clip_l.safetensors
in ComfyUI\models\clip
For GGUF version
I2V Quant Model : wan2.2_i2v_high_noise_14B_QX.gguf and wan2.2_i2v_low_noise_14B_QX.gguf
In models/unet
Quant CLIP: umt5-xxl-encoder-QX.gguf
in models/clip
FLUX_GGUF_Model : G8, G5, G4
in ComfyUI\models\unet
FLUX_CLIP : Q8, Q6, Q5, Q4, Q3
in ComfyUI\models\clip
For speed version
lightning LoRA : Wan2.2-Lightning_I2V-A14B-4steps-lora_HIGH_fp16.safetensors and Wan2.2-Lightning_I2V-A14B-4steps-lora_LOW_fp16.safetensors
For AIO version
I2V FAST AIO Model : wan2.2-i2v-rapid-aio.safetensors
In models/checkpoint
FLUX Text encoder : ViT-L-14-TEXT-detail-improved-hiT-GmP-TE-only-HF.safetensors
in ComfyUI\models\clip
VAE: wan_2.1_vae.safetensors
in models/vae
FLUX VAE : ae.safetensors
in ComfyUI\models\vae
ANY upscale model:
Realistic : RealESRGAN_x4plus.pth
Anime : RealESRGAN_x4plus_anime_6B.pth
in models/upscale_models
Description
What's new? :
ComfyUi 0.3.29 positive node fix,
LoRA fix with torch compile.
FAQ
Comments (35)
What is in this v2.1 update? Just fixes?
All in in "about this version" :
What's new? :
ComfyUi 0.3.29 positive node fix, (because last update break lot of node)
LoRA fix with torch compile. (now lora work with torch compile)
@UmeAiRT thanks i've been using this site for months and never noticed that section
Author just broke it 🤣 (I am joking) but version 2.1 not working for me
All your ComfyUI nodes are missing required inputs, like vae_name, expression, format, frame_rate, noise_seed, sampler_name, etc.
@DarkMaster13 If you have installed the mixlab nodes, delete them and see if the problem persists.
I'm really confused, I've used this before and had no issues, now I'm getting OOM and my settings everything is the same, am I missing something, I'm running 4060 8gbvram 32gb ram I've tried sage I've tried lowram. It will complete all the way up to the part going into making the video then I'll get the OOM. Before it didn't do that, I am using GGUF too.
Same, new workflow 2.1 make almost my GGUF OOM.
@ritonenier What i ended up doing is using face2video then bypassing the winvideo node so it stops, then i use the picture i just made into image2video, and that worked i guess for now
Using FLUX and WAN in the same workflow requires a lot of VRAM. It is better to make the image with FLUX first as suggested by drdewme618 to save VRAM. (torchcompile use lot of VRAM too)
@UmeAiRT Is there anyway to make a FACE TO IMAGE workflow? THen i could just use that face and put it into image2 video. if possible thank you for the reply
Add a Clean VRAM mode and Unload all models node after each generation. I do this and can run on 8GB VRAM laptop GGUF.
@mrw21j can you share your workflow? mine not working
This workflow causes issues after issue, ive fixed about 4 hours of issues and then 20 more crop up, rebuilt everything, reinstalled everything, and still, i feel like i need to get a team of code scientists to work on this 24/7 for a year just to backtrack all the missing dependencies.
Just use his all in one installer. You probably just messed up somewhere. Easy to do. Comfy, and python in general is finnicky as hell.
@johannesdolch539 his installer couldn't pass middle steps and just drops lots of errors then freezes the command line, was even trying to install it on a freshly installed clean windows, simply not working. And according to comments, there isn't just a single problem, unfortunately he broke it with updates
I'v got an issue with the following line in python 3.12.9
.\python.exe -m pip install --use-pep517 facexlibpython.exe -m pip install git+https://github.com/rodjjo/filterpy.git
return value "no such option: -m"
All the requierement seems well install, i've got the facexlib into this folder "ComfyUI_windows_portable-0.3.29\python_embeded\Lib\site-packages\facexlib\"
it 2 different commands use it separately
python.exe -m pip install --use-pep517 facexlib
python.exe -m pip install git+https://github.com/rodjjo/filterpy.git
and for python 3.12 (download insightface-0.7.3-cp312-cp312-win_amd64.whl before of cource:) )
python.exe -m pip install onnxruntime==1.19.2 onnxruntime-gpu==1.19.2 insightface-0.7.3-cp312-cp312-win_amd64.whl
No module named 'fused_layer_norm_cuda'
Same issue here.
@ag13trash994 This problem comes from apex which is not present or not compiled with CUDA
@UmeAiRT I used your most recent version of the all in one installer. It looks like it was not able to install apex into the embedded python at all.
@ag13trash994 I fixed this problem in the installer for FLUX, I have to do it for WAN too.
@UmeAiRT 10-4. Thank you
@UmeAiRT thanks, so i have to run the installer again?
@ag13trash994 thanks a lot
i have this issue how can i fix it
PulidFluxEvaClipLoader
No module named 'fused_layer_norm_cuda'
@ag13trash994 you installed the CPU version instead of GPU version, xxx_cuda is GPU version, since you renamed it it will use the CPU (slower)
getting the same error as well,
Also go the "fused_layer_norm_cuda" error. Fixed with the suggest rename, then I progress to this error:
ApplyPulidFlux
CUDA error: no kernel image is available for execution on the device CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect. For debugging consider passing CUDA_LAUNCH_BLOCKING=1 Compile with TORCH_USE_CUDA_DSA to enable device-side assertions.
I am using the combined installer, and get no issue with i2v workflow.
I got the same error
love this workflow! any chance you'll make it for wan2.2?
Same question
Download Insightface whl which corresponds here : Assets/Insightface
Good afternoon! Can you tell me where to download it?