[Edit:
Version v5.0 works with latest comfyui (v0.15.0).
If you have any problems, please refer to the FAQ at the bottom of the page or have a look in the comments.
Many thanks to everyone who tested this workflow. Thank you very much for the many inquiries and, of course, for all the knowledge and experience you have contributed. here👍🙂
Special thanks to:
@SeoulSeeker for the "Dead Simple MMAudio" workflow wich are the basis of the audio part here,
@taek75799 for the really well working enhanced models
@Bakazaya pointing to the color issue in version v3.0 and running lots of tests,
@bluntfeather sharing latest experiances with installing Comfyui-Easy-Install,
@nitrovtx for remain persistent in matters of quality and running a lot of tests,
@Icey64 for providing the link to "Comfyui-Easy Install",
@boinobin730 for asking for a First to Last Frame option, running pre tests and responding fast as hell 🙂 and
@SnowShoes311 thank you so much again for all your buzzing 😋]
Features:
Optimized Wan 2.2 workflow, runs perfect on RTX 3060 12 GB VRAM GPU and 32 GB RAM,
"Text to Video", "Image to Video" and "First/Last Frame 2 Video" generation in one workflow, all with easy audio generation,
easy installation/model downloading, all necessary sources are specified,
easy to use workflow, clearly structured, all necessary steps are explained,
easy switches for mode selection,
easy prompt selection for fast prompt creation/testing,
easy switching between "standard" and "enhanced" models,
very fast and smoth high quality outputs up to aprox. 1440 x 960 with 60fps,
2x fast upscaler,
4x fast framerate multiplier,
MMAudio Sampler (generates sound accordingly to the video action),
Triton and Sage Attention option,
A 5 Second long high quality video generation takes about 10 - 15 minutes (see below).
Tested generation times:
As a rough guide value for RTX 3060 GPU: generating a 5 second long high quality 1440 x 960 60 fps video with 6 steps it will take:
t2v: around 10 - 12 minutes,
i2v: around 15 minutes.
Comfyui-Easy-Install with Triton + SageAttention:
This workflow should work with any latest comfyui version >v0.6.0 (Desktop, Embedded, Windows/Linux).
However, comfyui is developing rapidly, and it often happens that some of the custom nodes used are not updated quickly enough or not updated at all. Manual workarounds are sometimes necessary. Furthermore, care must be taken to ensure that there are no conflicts with other nodes.
If you're having difficulties with your existing comfyui system or if you want to run video generation on a separate (parallel) comfyui system, like I do, I would recommend you the following installer: https://github.com/Tavris1/ComfyUI-Easy-Install.
Complete installation of comfyui including manager and some pre configured custom nodes is just one click - really 🙂
Installation of Triton + SageAttention is just a second click - really 🙂 And since it's so easy now, I would definitely recommend it to you for video generation.
Cause it is an embedded version, you can install it parallel to your existing comfyui version without the risk to ruin your working system.
After installation just configure the "extra_model_paths.yaml" file to use your existing models.
After a fresh installation of Comfyui-Easy-Install you might have some issues too, but there are known workarounds - please see the FAQ below.
For testing/understanding/experimenting/changing the workflow:
Click "Toggle Link Visibility" to see the links.
click the Subgraph symbols to open the Subgraphs.
for quick testing you may lower the settings for: steps, clip lenght and video resolution,
be really carefull with modifying Groups or Subgroups (even Titel or Color) cause they are essential for switching,
feel free to try and test other models. Just give me a hint if you find models which deliver better results and fitting the 12 GB VRAM limit.
And as usual: Have Fun 🙂🙂
Short Conclusion:
This workflow is based on elements of a variety of allready published workflows. My "job" was only to put things together, optimize it for a small machine and create a most simple and hopfully user or even "beginner" friendly workflow.
I`m not an "expert" - just a user who wants to get it running on "available" hardware.
There are many things I don't really understand. If you find mistakes or better solutions please give me a hint.
And I really hope that even "beginners" have a chance to go the first steps...
Frequently Asked Questions (FAQ):
For quick and better overview I will try to merge all known issues here - step by step (please be patiant). If your issue is not listed here, please have a look in the comments first. Most issues have been allready discussed.
Comfyui Nodes 2.0:
Turn off Nodes 2.0 in comfyui (use comfyui menue). Actually not all custom nodes are supported.
Comfyui crashes after generation while vae decode, upscaling or frame rate multiplying (Rife VFI) without any error report:
This is a RAM problem (not VRAM). Increase your swap file (min. 64 to 128 GB) or set it to automatic management on a fast drive with at least 100 GB free space.
JW Nodes (JWFloatToInteger, JWIntergerDiv, JWImageResizeByLongerSide), soundfile missing:
For the workaround look here and here:
python -m pip install soundfileFresh Comfyui-Easy_Install Installation (missing soundfile and Pytorch v2.9.0 issue with SageAttention on Windows:
For full conversation look here.
Open cmd in python_embedded folder:
python -m pip install soundfile python -m pip uninstall -y torch torchvision torchaudiopython -m pip install torch==2.8.0 torchvision==0.23.0 torchaudio==2.8.0 --index-url https://download.pytorch.org/whl/cu126Slider Nodes - how can I modify the "default" values:
Right click the slider node, choose Properties and set the values you like 🙂🙃
Description
Workflow adjusted to latest comfyui versions,
Improved KSampler control (generation steps now splitted over High and Low Pass),
Slightly faster generation (5 instead of 6 steps for otimal outputs),
Improved First Frame Image to Video generation for better movements (PainterI2V node),
Significantly better I2V movements by using "enhanced" models,
Easy switching between "standard" Q4 models and "enhanced" Q8 models.
Some "design" and "readability" improvements by using Subgraphs.
FAQ
Comments (38)
Here at I2V it's stuck at 720, but I wanted to increase it to at least above 900
@paulo Hi - please provide usefull information. No one will solve your issue: "it stucks" 😉
Wich workflow version?? What "stuck"??? Wich node? VRAM, RAM???? Errors???? Just a crash without errors?? Did you allredy read my brand new FAQ?? 🙂
@arkinson I think the point is that maximum resolution is 720p, but he wants to generate 900p and even bigger 😅
@Bakazaya Oh bro - I believe I need a "slang" translator at my last old days 🤣
Ok - here we go: Right click the slider node, choose Properties, set the values you like 🙂🙃
@arkinson Sorry for the confusion, guys, I used Google Translate rsrs
@paulo No, no - thank you (or google) for the fun 🤣 I just released the new workflow version and was completely fixed at complicated questions about new bugs. What do you call something like that - operational blindness? 🙄🙂
Can't get this to work, tried multiple versions including the latest and *comfyui desktop* fails to install *custom nodes* every time. I'm skipping the models because I already some that should work with the flow but the custom nodes always fail to install. I am new to this but I guess some things don't just work like they appear like they should?
*edited for clarity
Hi - I`m not sure if I understand you right. What do you mean with versions - Comfyui Desktop/Portable/Standalone??? wich operating sytem? Or do you mean latest release??? What means: "the dependencies always fail"???
If you are a "beginner" and you don`t have a working comfyui system I really would recommend you a fresh installation with "Comfyui-Easy-Install", just have a look in my brand new model description for detailed information. The completely automated installation will take you just 2 mouse clicks and around 30 minutes of time.
After installation just configure the "extra_model_paths.yaml" file to use your existing models. Then load my latest workflow version, install all missing nodes via the manager, load and check all models as discribed and it should work out of the box without any issues 🙂
I'm on Windows, Comfyui Desktop app.
Sorry: I meant to say I tried different version of this workflow, including the latest release. On comfy .7 which I also installed today. The dependencies for the workflow weren't installing regardless of trying different versions of your workflow and restarting comfyui to try again. I have comfyui working with another civitai workflow so it seems to be in a functional state but yeah, dependencies are saying they are installed when they aren't on this one. Probably I'm doing something wrong, thank you I may try that if all else fails.
I went back in and looks like it partially installed the extensions/dependencies, but does not ever complete the install of the following:
PainterI2V
JWFloatToInteger
JWIntegerDiv
JWImageResizeByLongerSide
FinalFrameSelector
Will see if it's feasible to try to add those myself but still appreciate any help thank you
@bluntfeather You did not anwer my questions and we actually are on comfyui 0.8.2. What are comfyui "dependencies"?? You mean custom nodes - or what are you talking about? Sorry, but without usefull informations I can`t help you.
@arkinson Sorry, it was .7 yesterday but updated to ComfyUI_desktop v0.7.2 today. No change in the custom nodes not working (I was wrongly referring to them as dependencies)
Basically after the pop up where it would have you download models, it has another pop up window where you'd download the custom nodes necessary to run the workflow. These nodes are not installing properly:
PainterI2V
JWFloatToInteger
JWIntegerDiv
JWImageResizeByLongerSide
FinalFrameSelector
@bluntfeather Sorry, I overread "I'm on Windows, Comfyui Desktop app." Thank you. I run it on this version too, just with the JW nodes workaround. So, your other issues seems to be node conflicts, as mentioned.
@arkinson Thank you, will try that easy-install method you mentioned (take it it's this: https://github.com/Tavris1/ComfyUI-Easy-Install), happen to know if since that appears to be portable I could potentially have both the official comfyui desktop and this comfyui-easy-install both installed at the same time as long as I only ever have one running? Not sure there's a reasonable use case for it but just wondering. Thank you for your help and for providing your workflow!
@bluntfeather Yes, that is the correct version and yes - install it in parallel. As mentioned - have a look here in my model description - I allready wrote a more detailed guide. The most advantage of these embeded versions is - you get a perfect test invironment wich is absolutely usefull for the fast developing AI. We are on comfyui 0.8.2 now 😉🙂
@arkinson I did the comfyui-easy-install method and it went more smoothly but still had an issue with two of those extensions. The error log had this in it: ModuleNotFoundError: No module named 'soundfile'
So I went to the python_embedded folder in CMD and entered:
python.exe -m pip install soundfile
Reset my server and your worksheet loads now, woot. Baby steps
@bluntfeather Thank you for your feedback👍 I`m glad you got it running and happy generating 🙂
@arkinson Thank you for the help! Also ended up having an issue with sageattention when running the flow as an error occured installing it with comfyui-easy-install, had to downgrade to pytorch 2.8 as it seems there's no sageattention yet for 2.9 on windows..
@bluntfeather Ah - thank you for the information. I completely updated my existing system yesterday and I´m still on Pytorch Version 2.8.0+cu128. How did you perform the downgrade? I would like to publish this in the FAQ.
@arkinson
Sage version I installed: sageattention-2.1.1+cu128torch2.8.0
Torch:
python.exe -m pip uninstall -y torch torchvision torchaudio
python.exe -m pip install torch==2.8.0 torchvision==0.23.0 torchaudio==2.8.0 --index-url https://download.pytorch.org/whl/cu126
And not sure if this is helpful for the faq, but below I include instructions for downloading and placing all the models/vaes/etc the workflow uses. I loaded all your models to see if it'd "just work" after getting it to load properly and correcting the sageattention and torch stuff, and got that old dude generated just fine, the only difference I believe are there is an addition low noise file where you were using one for both high and low noise before, think it was the bouncy walk one:
1. Standard Wan2.2 GGUF Models (4 Files)
These are the core video generation models. You need both the "High" and "Low" noise versions for both Text-to-Video (T2V) and Image-to-Video (I2V) modes.
Repo: Hugging Face - Wan2.2-T2V-A14B-GGUF
File 1: wan2.2_t2v_high_noise_14B_Q4_K_M.gguf
File 2: wan2.2_t2v_low_noise_14B_Q4_K_M.gguf
Repo: Hugging Face - Wan2.2-I2V-A14B-GGUF
File 3: wan2.2_i2v_high_noise_14B_Q4_K_M.gguf
File 4: wan2.2_i2v_low_noise_14B_Q4_K_M.gguf
2. Enhanced Wan2.2 GGUF Models (2 Files)
These are the custom "NSFW/Enhanced" versions used in the "Enhanced Models" switch.
Page: Civitai - Uncensored WAN 2.2 Remix V2
File 5: wan22EnhancedNSFWCameraPrompt_nsfwFASTMOVEV2Q8H.gguf (High)
File 6: wan22EnhancedNSFWCameraPrompt_nsfwFASTMOVEV2Q8L.gguf (Low)
3. LoRA Models (4 Files)
These control the motion and style.
Lightning Distill: Hugging Face - Wan2.1-I2V-14B-Lightx2v
File 7: Wan21_I2V_14B_lightx2v_cfg_step_distill_lora_rank64.safetensors
Bouncing Boobs: Civitai - Wan Bouncing Boobs
File 8: wan Bouncing Boobs i2v 14b v1.safetensors
Bouncy Walk: Civitai - Wan2.2 Bouncy Walk
File 9: wan22 Bouncy Walk v2 Low.safetensors
File 10: wan22 Bouncy Walk v2 High.safetensors
4. Support Models (2 Files)
These are for the upscaling and smoothing (frame interpolation) steps.
Upscaler: Hugging Face - Real-ESRGAN
File 11: RealESRGAN_x2.pth
Frame Interpolation: Hugging Face - RIFE Models
File 12: rife49.pth
----------------------------------------------------------------------
Here is the exact map of where to drop each file inside your main ComfyUI folder:
1. The UNET Models (6 Files)
Place these directly in: ComfyUI/models/unet/
wan2.2_t2v_high_noise_14B_Q4_K_M.gguf
wan2.2_t2v_low_noise_14B_Q4_K_M.gguf
wan2.2_i2v_high_noise_14B_Q4_K_M.gguf
wan2.2_i2v_low_noise_14B_Q4_K_M.gguf
wan22EnhancedNSFWCameraPrompt_nsfwFASTMOVEV2Q8H.gguf
wan22EnhancedNSFWCameraPrompt_nsfwFASTMOVEV2Q8L.gguf
2. The LoRA Models (4 Files)
Important: You must create specific sub-folders inside models/loras/ because the workflow looks for them there.
Create folder: ComfyUI/models/loras/91 Lighning/
Drop this file here: Wan21_I2V_14B_lightx2v_cfg_step_distill_lora_rank64.safetensors
Create folder: ComfyUI/models/loras/90 video/
Drop these 3 files here:
wan Bouncing Boobs i2v 14b v1 (her breasts are bouncing).safetensors
wan22 Bouncy Walk v2 Low.safetensors
wan22 Bouncy Walk v2 High.safetensors
3. The Upscaler (1 File)
Place this in: ComfyUI/models/upscale_models/
RealESRGAN_x2.pth
4. The Frame Interpolation (1 File)
This usually goes into the custom node's own folder. Check this path: ComfyUI/custom_nodes/ComfyUI-Frame-Interpolation/ckpts/rife/
rife49.pth (Note: If that folder doesn't exist, try creating it, or check ComfyUI/models/frame_interpolation/ depending on your version).
@bluntfeather Hi - thank you so much for the pytorch downgrade commands and the link 👍 I will add your hints for a fresh Comfyui-Easy-Install to the FAQ soon.
The model download links and save directories are all specified in the workflow itself. Users need just to click on the download links and manually move the models to the specified directories.
Btw. there is no need for "special" subdirectories (extra path yaml should be configured correctly of course). The main point is, the user has to check/select "his own local models" in every loader node.
@bluntfeather FAQ edid is done 🙂 Thanks again.
@arkinson Cool! Just one note, where you have the pytorch info in the FAQ, probably want to try to keep the url looking less like a separate link and more like part of the command, i.e. python.exe -m pip install torch==2.8.0 torchvision==0.23.0 torchaudio==2.8.0 --index-url https://download.pytorch.org/whl/cu126
This site likes to auto-link url's but it can be disabled per-url basis, like I did above. Just select the whole link and a ui popup will let u unlink it
@bluntfeather Oh my 🙄 You are right. I just copied your comment. I hate this civitai text editor... Fixed now. Thank you for reviewing. And have a look at my new workflow version v3.1 🙂
@arkinson had some issues trying the image to video route, tho I have 32gb of ram kept having one memory issue or other. after swapping the vae decode node (inside upscale + rife) with vae decode (tiled) and changed the following.
RIFE VFI Node: Set multiplier to 2 (not 4).
Set clear_cache_after_n_frames: 2
VAE Decode (Tiled): Set temporal_size to 32 (not 64).
Which worked a bit but seemed to need to restart comfyui every time to do a gen. So for the standard gen unet model I swapped out your wan2.2 itv with RapidWan2.2, bypassed both additional LORAs, and also increased my paging file from its automatic 42gb~ setting to manual 32-64gb setting. Still tinkering with somethings like Video Combine Final (30fps too slow, 60fps seems slightly fast but only two seconds long, may set that rife multiplier back to a higher value...) but it's running everytime now. Think that's all I did lol, didn't notice any difference in output quality but very limited testing.
@bluntfeather Had you looked in the FAQ?? 😉 The way you went is possible when you run out of RAM (not VRAM) (tiled vae and lowering clear_cache_after_n_frames), but increasing the Windows swap file is the most easy way if you have enought free space on a fast drive. We had this "issue" very often in the comments - increasing swap file allways solved it. Look in the FAQ for more detailes.
@arkinson woops! I tried your new version, so far all good no issues, like the extra tips :D
I tested new PainterI2V node with values of 1.3 and 1.5 and found that it only works properly with "enhanced" models. With "standard" models, there's a noticeable color change at 1.3 value. At 1.5, the colors shift dramatically, and sometimes even the background becomes abstract something. Here's an example (first is "enhanced" with value of 1.5, second is the "standard" model with 1.5) : https://civitai.com/posts/25733267
@Bakazaya Hi - thank you so much for first testing and pointing out the color issue. I just saw, there is a brand new updated versions of the PainterI2V node: https://github.com/princepainter/ComfyUI-PainterI2V --> https://github.com/princepainter/ComfyUI-PainterI2Vadvanced wich seems to solve the issue. It could be installed with the manager. Let`s test it 🙂
@Bakazaya Adding the PainterI2V advanced node was a little bit tricky, caused by the switch options, but at least the I2V part should work. If you like, try my testflow "wan v3.0 test painter advanced.jason" from here: https://civitai.com/articles/23042.
@arkinson Yeah, I tried to connect it myself, but I couldn’t figure it out at all, lol, so I’ll try yours😁
@arkinson I ran a few tests and didn't notice any color issues with 1.3 value. It seems to change color slightly with 1.5, but i'm not sure, so more testing is needed. I also want to check how much this affects prompt following. It feels like results are more random with 1.5 value. I'll do more testing tomorrow. But Advanced Node is better then default PainterI2V anyway.
@arkinson I spent several hours on tests. No problems were noticed with "enhanced" models. With "standard" models, in most cases everything is fine, both at 1.3 and 1.5 values. Changing color and background is still possible, but it happened to me in one of 20 generations. I also did not notice a significant effect on prompt following, even at 1.5 value. My verdict is that "PainterI2V advanced" works almost perfectly compared to the regular version.
@Bakazaya Thank you so much for your tests 👍 I haven't had time yet. But your results sounds very good. I will do some tests, if anything else is working, because the switch logic is sometimes a bit malicious. I will do a short description/note about the setting and try to arrange some "minimally invasive standard" settings. Hopfully all this stuff did not overhelm the "common" user....
@Bakazaya Ok - version v3.1 is out now and seems to work in all cases. Thanks again for well usable input.
@arkinson I've already tested it a bit, so far everything seems to be ok 👍
@Bakazaya You are faster then lightspeed, thank you 🙂