⚡ Recommendation
You can use my custom checkpoint V2 NSFW, Lightning included:
➡️ WAN 2.2 Enhanced – Lightning Edition
for wf lightningedition long v2.1
https://civarchive.com/models/2053259?modelVersionId=2477539
Or for wf svi , no Lightning included:
➡️ SVI NSFW
https://civarchive.com/models/2053259?modelVersionId=2609141
Update: 2 or 3 KSamplers
The triple-KSampler workflow added dynamism to avoid slowdowns, but it could cause issues in certain sections, such as high brightness or other bugs. To fix this, simply enable the Double KSampler in the problematic section and change the seed.
🟣 uptade SVI LoRAs – Strengths & Weaknesses
⚡ Overview
✅ Strengths
Best solution for making long videos
Excellent transitions between video segments
Reduced degradation compared to other solutions
Strong character coherence: the model retains information from the previous video, helping maintain consistency
⚠️ Weaknesses
Weaker prompt understanding
Weaker camera understanding
Videos are less dynamic
Sometimes slow-motion effect
(can be improved with proper Lightning LoRAs and dynamic prompts)
🟣 SVI LoRAs – Download Links
⚡ Download
Note: Both LoRAs must be loaded manually in your workflow.
🟣 Suggested Lightning LoRA Combos (Optional)
⚡ Overview
You don’t have to use these Lightning LoRA combos. They are optional and allow you to fine-tune motion and degradation.
You can also use other Lightning LoRAs or assign different combos per video for more control.
🔥 Combo 1 – More Motion (Rapid Video Degradation)
High LoRA →
https://huggingface.co/Kijai/WanVideo_comfy/blob/709844db75d2e15582cf204e9a0b5e12b23a35dd/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank128_bf16.safetensors
Weight: 4
or
💜 Combo 2 – Less Image Degradation
or
or
🟢 Combo 3 – Balanced Motion / Moderate Degradation
High LoRA →
https://huggingface.co/Kijai/WanVideo_comfy/blob/709844db75d2e15582cf204e9a0b5e12b23a35dd/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank128_bf16.safetensors
Weight: 3
Low LoRA →
https://huggingface.co/Kijai/WanVideo_comfy/blob/709844db75d2e15582cf204e9a0b5e12b23a35dd/Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank128_bf16.safetensors
Weight: 1.5
If you want to create several-minute-long videos while maintaining high quality, it is possible, but it will take a very long time. You just need to avoid using Lightning LoRAs and use the full model instead.
🧠 Advanced Usage Tip
You can disable global Lightning LoRAs and assign different combos per video:
Combo 1 for Video 1, Combo 3 for Video 2, Combo 2 for Video 3.
Each combo produces different motion and degradation behavior.
If you don’t use Lightning LoRAs, degradation will be even lower, but generation time will be very long.
🧠 Dynamic Prompts – Better Control & More Motion
You can use dynamic prompts to have better control and this helps to make the video more dynamic.
You just need to give this example prompt to an LLM like ChatGPT. It will be enough for it to describe the image you have and what you want as a video while keeping the same structure of the prompt in the following example.
⚠️ This will be a NSFW prompt; ChatGPT will not accept it.
You can use GROK https://grok.com, which can make NSFW prompt modifications.
For more examples of prompts with different poses, check:
Enhanced FP8 Model.
Give these prompts to GROK with the dynamic prompt structure if you want.
Example 1
(At 0 seconds: Wide shot showing a slightly overweight man casually walking down a city street, camera fixed in front, urban environment with buildings and cars.)
(At 1 second: Suddenly, a massive shark bursts from the pavement ahead, looking terrifying at first, pavement cracking, dust and debris flying, camera from side angle.)
(At 2 seconds: Medium shot from the side, the man stumbles backward in shock, while the shark dramatically slows down and strikes a comically exaggerated sexy pose, revealing large, exaggerated shark breasts, covered by a colorful bikini.)
(At 3 seconds: Close-up on the man’s face, eyes wide in disbelief, as he turns to look at the shark, small cartoon-style hearts floating above his head to emphasize his amazement, camera slightly low-angle.)
(At 4 seconds: Dynamic travelling shot showing the man frozen in the street, the shark maintaining its sexy pose, water splashes and debris still moving realistically, urban chaos around.)
(At 5 seconds: Wide cinematic shot pulling back, showing the man standing in the street, staring at the bikini-wearing shark with hearts above his head, epic perspective highlighting absurdity and humor.)
Example 2 – Anime NSFW
(At 0 seconds: The couple in a cozy bedroom, anime style, soft lighting highlighting their intimate embrace, her back arched slightly as he positions himself.)
(At 1 second: The man’s hips moving rhythmically, the head of his penis sliding effortlessly into her vagina, her body responding with a gentle, fluid motion, anime-style motion lines emphasizing the smooth penetration.)
(At 2 seconds: Her back arching deeply against him to intensify the pleasure, hips swaying with each thrust, breasts bouncing subtly, small hearts floating around them to capture the erotic energy.)
(At 3 seconds: Her face, eyes closed in bliss, a soft moan escaping, hands resting behind her head, anime-style blush on her cheeks, the air filled with a seductive aura.)
(At 4 seconds: The man penetrating her deeply, her body moving in sync with his, the bed sheets slightly rumpled, the room’s warm lighting enhancing the intimate, lustful atmosphere.)
(At 5 seconds: The couple locked in a passionate embrace, the scene exuding vibrant, seductive energy, anime style with smooth lines and soft shadows.)
findperfectresolution
https://github.com/ashtar1984/comfyui-find-perfect-resolution
rgtree
https://github.com/rgthree/rgthree-comfy
kj-node
https://github.com/kijai/ComfyUI-KJNodes
iamccs
https://github.com/IAMCCS/IAMCCS-nodes
🔧 What’s New – V2.1
The workflow is now much simpler than before.
You can create videos up to 20 seconds.
New sections can be added very easily by copying and pasting an existing section.
🎨 New Node: PainterI2VAdvanced
🔗 https://github.com/princepainter/ComfyUI-PainterI2Vadvanced
Main Parameters
color_protect
Enables or disables color drift correction.
Recommended: True.
Prevents green tint, darkening, and color distortion.
correct_strength
Controls the strength of color correction.
Higher value = stronger color correction
findperfectresolution
https://github.com/ashtar1984/comfyui-find-perfect-resolution
_______________________________________________________________________________________________
New FunVace workflow: it's extremely easy to use. If you have separate videos and want to join them with a seamless transition, use this! Thanks to Bob once again, excellent work: https://civarchive.com/models/2277993?modelVersionId=2563927
Workflow Update – Lightning Edition V2
📌 New Version Available
The Lightning Edition V2 is the latest update of my workflow.
This version is cleaner, better organized, and scene transitions are improved for smoother results.
🔧 Fixes & Improvements – Lightning Edition V2
Workflow fully reorganized and clarified for better readability.
Scene transitions improved for smoother results.
Optimized nodes and settings for easier use.
Confirmed compatibility with the Lightning Edition V2 model.
🔹 V1
Version without Super Power LoRA
A version without Super Power LoRA has been added for those who are more cautious.
If you want to try the Super Power LoRA without using the folder included in the ZIP, check the GitHub page:
➡️ GitHub – rgthree workflow
Tutorial
The workflow is divided into 6 sections:
Section 1: generates a 5-second video.
Section 5: handles transitions and interpolation.
Important notes:
Running only sections 1 + 2 will not merge the video.
To create a 15-second video, run sections 1 + 2 + 5.
Interpolation can also be enabled in section 1 if you prefer.
You can use different prompts and LoRAs for each section.
Resolution is automatically adjusted to avoid cropping.
💥 Super Power LoRA – V1
Requires the custom rgthree node:
➡️ rgthree ComfyUIInstall via git clone or the Manager, then replace the folder with the one included in the ZIP.
Allows you to save and load templates, very useful when working with multiple LoRAs, especially for WAN 2.2.
A version without Super Power LoRA is also available for those who prefer.
For the version with Super Power LoRA, see: GitHub – rgthree workflow
✅ Tips – V1
Keep the default settings for fast results.
Increase the steps if you want better quality.
Description
FAQ
Comments (70)
Is there any way to utilize the PainterI2V custom node with Kijai's WanWrapper workflow?
PainterI2V makes your custom model even better. With it you can eliminate all the usual slow motion for I2V. Any suggestions for a T2V equivalent? Now it's T2V that feels slow.
Impressive workflow with outstanding results and user-friendly. By the way, I often encounter inconsistent facial features of characters during use. If possible, could you add IPAdapterWAN in future versions (if any) to maintain a consistent appearance? Here's an example: https://civitai.com/models/2005198/lazy-wan-22-i2v-workflow
thank you for your excellent work. Looking forward to your reply.
I wanted to try your workflow, took a look at it and closed it immediately. Its a mess with all the spaghetti links. It might work for you, but please, if you make a workflow for the community, make it clean and organised.
For me somewhat the results are pixelated/blurry or with plenty or artefacts. tried to use more steps, less crf, higher resolution (for example 800x800), but it still was super pixelated or blurry. using a 2000x2000 image as input. also activated "upscale" for test purposes but didnt change anything
Thanks! Giving this a try now. The two custom nodes that manager and I have failed to install so far are Super Power Lora Loader (is this because I already have Power Lora Loader from rgthree?) and SimpleMathInt+
Ah I see your note about super power lora loader and github, I'll look at that.
@imafan927 If you take the version with the Super Power LoRA (named V1) in the ZIP, there’s the rgthree folder. You need to replace this folder with the rgthree folder in your custom node directory. The Super Power LoRA will then appear.
But it’s not mandatory, you can also just use the Power LoRA.
@taek75799 Thanks!
@taek75799 how to install the node SimpleMathInt+?
@mmikemiller823390 So im having the same issue with the supermathint+ node. its part of the comfyui essentials pack and when i look in the comfyui manager the node is listed but i have no explanation as to why it does not show up for my WF either. Regardless, that issue is not on @taek75799, but i wish i had a a solution as well. @taek75799 Thanks for that super power lora loader, love it!
@mmikemiller823390 @barnold946329 Hello, the SimpleMath node is not mandatory, it just makes things easier. You enter a resolution for the width and height. The most important one is the Find Perfect Resolution node, which gives you the correct video ratio that matches the image. If you want to replace that node, you can use this one from Kijai’s custom: https://ibb.co/2YShSJsR
hello and thanks for your work here,
may we use your workflow to combine your model base with any lora sd 1.5 for Wan 2.2?
Unfortunately, no, it’s impossible. SD 1.5 LoRAs will not work on WAN. Only WAN 2.2 LoRAs and some from WAN 2.1 will work.
@taek75799 ok right, thanks for your answer
@dreamswalker Check the next thread, would you also mean a mistake in the card?
@qek simply the lora doesn't work
but I have tried to use the suggested Cliploader and got an issue
the problem looks like hat file is not in the WAN “repack” format for Comfy
maybe it is a “pure” text-encoder (intended for diffusers or for a custom setup by the creator), without the tokenizer piece that Comfy’s CLIPLoader expects, maybe it's for A1111
I'm not sure, in any case for me the custom Cliploader didn't work
The latest version has Base Model: SD 1.5 in the card, why?
Hello, and thank you very much. I must have made a mistake; I’ll correct it.
Hey! what am i supposed to put in the power lora node?
The light loras? High and low? Cause the results get somewhat blurry and noisy.
Thank you!
Hello, it depends on the model you’re using. If you’re using my model, you don’t need any additional Lightning LoRAs, otherwise the video will be degraded. If you use the native Wan model, then yes, you need the Lightning LoRAs if you want to generate your videos in 4 steps.
@taek75799 Thank you for the response, i will try it out <3
I downloaded your workflow and model but i'm still confused on to how to use your model within the workflow, could you give me some hints? thanks in advance
Hello, it's quite simple. In the Load Diffusion Model node, on the left, the top one must load the high model, and the one below must load the low model. For WAN 2.2, you are required to load both.
@taek75799 I got it working now, thanks!
how to install super power lora loader?
Hello, you need to install the RGTRee custom node first. Then download the v1 of the workflow from here. Inside, there’s an RGTRee folder—go to your custom node folder and replace it. The Super Power LoRA should then appear.
What measures can be taken when insertion and removal operations are fast?
If you have any ideas, I'd appreciate hearing them.
I don’t think withdrawal is possible, even if you specify it in the prompt. You would need to create a LoRA for that.
For insertion, I recommend this LoRA, for example with a prompt like:
‘The man holds his erect penis, then slowly inserts it into her vagina.’
https://civitai.com/models/2100959?modelVersionId=2377244
I had three questions implementing your workflow. One, the two diffusion models don't match the names of any files on your main download page, and I'm not good enough to identify which ones are appropriate. Two, what should go in the section 1 prompt vs section 2, etc. Three, it's not clear to me how one sets the video length.
Hello, don’t worry about the model name: Civitai changes it automatically when I upload it, so it won’t be the same as mine. Check the description section on my model page and pick whichever you prefer.
For the prompt in the section, put what you want to see during the first 5 seconds of the video, since the number of frames is fixed at 81 this is the maximum WAN can handle; above that, it often produces poor results.
For the second prompt, put what you want to see for the next 5 seconds, and so on.
The video length is fixed at 81 frames; you can reduce it if you want. It’s located in → ImageToVideo under “Length”, just below Load Image.
Hi I have some problems here. When running the workflow with your Q6K model, the first run is fine, but when running for the second time, the output video is just noises, and if I keep running the workflow it always ends within two seconds and output fully noisy videos. What could be wrong?
Hello, could you send me the workflow along with the generated video?
You can use Pastebin, or this link if you’re not sure how to do it it’s easier: https://www.swisstransfer.com/fr-fr
Don’t worry, a lot of people ask me for help and everything works fine afterward.
@taek75799 thanks for the reply, I just DMed you
this is happening to me as well, first video gen was fine, anything else after is just full noise.
what was the solution?
@scooper86326 Hi! Try sending me the workflow you're having this issue with, and I'll check it out: https://www.swisstransfer.com/fr-fr
Very good workflow, when you are using non-local ComfyUI for generation the online service you use may experience troubles with rgthree nodes, but you can safely just bypass/delete them if you are not planning to use any loras. Also, double check if you have both models H and L correctly placed. That may save you a couple hours of debugging.
rgthree doesn't support Nodes 2.0
@taek75799 tryin to use ur workflows(including the ones in ur video samples)but keep getting this:
Command '['M:\\AI\\SwarmUI\\dlbackend\\python_embeded\\Lib\\site-packages\\triton\\runtime\\tcc\\tcc.exe', 'X:\\temp\\tmpvi7j31ab\\cuda_utils.c', '-O3', '-shared', '-Wno-psabi', '-o', 'X:\\temp\\tmpvi7j31ab\\cuda_utils.cp312-win_amd64.pyd', '-fPIC', '-D_Py_USE_GCC_BUILTIN_ATOMICS', '-lcuda', '-lpython312', '-LM:\\AI\\SwarmUI\\dlbackend\\python_embeded\\Lib\\site-packages\\triton\\backends\\nvidia\\lib', '-LC:\\Program Files\\NVIDIA GPU Computing Toolkit\\CUDA\\v12.8\\lib\\x64', '-IM:\\AI\\SwarmUI\\dlbackend\\python_embeded\\Lib\\site-packages\\triton\\backends\\nvidia\\include', '-IC:\\Program Files\\NVIDIA GPU Computing Toolkit\\CUDA\\v12.8\\include', '-IX:\\temp\\tmpvi7j31ab', '-IM:\\AI\\SwarmUI\\dlbackend\\python_embeded\\Include']' returned non-zero exit status 1.
Hello, did you enable the model patch torch setting, and patch attention nodes?
If so, disable them with Ctrl + B.
@taek75799 no they were disabled. let me give that a shot to see what happens
@emotionaldreams4 I don’t know SwarmUI very well, but I assume you can download the workflow that produces this error.
If so, could you send it to me? I know it’s mine lol, but many people contact me by DM for help, and often there’s a setting that I think was modified by mistake.
If you want, you can send it via this link:
https://www.swisstransfer.com/fr-fr
or Pastebin if you’re familiar with it.
@taek75799 the workflows are the ones you have up. i didn't change anything. have the same models used too. since i already had them. all i change was the image used and the prompt. also Swarm has Comfy portable imbedded in it. its on a tab
@emotionaldreams4 So the high model is indeed on the top load diffusion model, and the load diffusion model low is the one at the bottom.
Hmm, I don’t see where the problem is. It’s really hard for me to diagnose this remotely, especially with SwarmUI.
@emotionaldreams4 The error is related to Triton, so Sage Attention, and this error is caused by the node mentioned above, but they are disabled, so I don’t understand.
@taek75799 yeah triton..never had an issue with it til now
should I enabled all your workflow or just leave it be?
Hello, enable the 5th part if you want to get a 10-second video.
I invite you to check the description for more details. If you don’t understand something, feel free to ask here.
@taek75799 Yes I'm working out with your workflows. In parallel, I'm also working out with other design, if you please, would you see and check my workflow (I'll DM you if you don't mind)? Because my result it's not that good.
@nessusprod Of course, I don’t mind.
Do it whenever you want.
Any reason there is no sage attention node or patch torch settings node in the painter version?
Hi, I removed it. Some beginners tended to enable it even though it wasn’t installed on their system. I think people who know how to install it will know how to add the nodes. Everything works fine with Painter.
Hi guys, is there any solution to avoid reloading the two checkpoints during generation? I have 12GB VRAM and 32GB RAM, and I'm wasting a lot of time loading the checkpoints.
If they models don't fit into RAM at the same time, they will be unloaded. It's simply a matter of space. If they do, then look for 'clear cache nodes', but I suspect the total size of both is more than 32GB.
@Cochese9000 These are two 15GB Q8 models. I did what you said about "clearing the cache" and it really reduced the wait time a bit, thanks man!
Hi, you can try this workflow by @ApexArtist1 , it allows you to run GGUFs: https://civitai.com/models/2245346/wan-22-enhance-model-gguf-workflow. The Super Power LoRA isn't essential; it's not included in the latest version of the Painter-I2V workflow. Here is the link for Find Perfect Resolution: https://github.com/ashtar1984/comfyui-find-perfect-resolution
@taek75799 If I use the workflow you recommended, what about your workflow?
@cawaokaito I'm not asking anyone to use my workflow, I'm just sharing it. You can also use my workflow with GGUFs: just download "comfyui GGUF" from the ComfyUI Manager and replace the "Load Diffusion Model" nodes with "UNET Loader GGUF
@taek75799
The UNET loader GGUF cannot be loaded from the workflow you provided. If anyone encounters problems, it's recommended to download it from https://github.com/city96/ComfyUI-GGUF.git But I tried to fix this because I wanted to use your workflow; it's interesting.
Hey! Quick question about your workflow really love it
You mentioned the first generated clips can be upscaled to 60fps, but I noticed the rest of the video after the first 54 seconds looks closer to 16fps. How do you usually handle that?
how Do you i convert the generated parts to 60fps, or is there a way to make the whole video 60fps without it looking weird? Just trying to understand the best way to keep everything consistent.
Thanks!
Hello and thank you for your question. Actually, it's simple: first, there is the 1st section where you can interpolate your video to 60 fps, for example. In the second, you can't. Why? Because the videos are merged together in the last section (allowing interpolation for more than 5 sec), which is the 5th one.
If you want to make a 10-second video, you only need to activate the 5th block. If you want to make 15 seconds, you must activate the 2nd and the 5th. Keep in mind that only the 5th block allows for joining the videos. Also, the longer the video, the more it will degrade; let's say 15 seconds is still okay. Personally, I don't work this way anymore as I tend to use other tools. Unfortunately, it's not automated, and I have to switch between several interfaces. Here is an example using inpaint, upscale, the basic First Last Frame workflow, and my own workflow: https://civitai.com/images/114733916
@taek75799 Hey, thanks so much for the detailed explanation! Really appreciate it.
Just to clarify, I realized I made a mistake in my earlier message (meant 5 seconds, not 54, sorry about that!).
What I’m really trying to figure out is how to get that smooth 60fps effect throughout the entire 15 seconds of the video, not just the initial part. Are you using any additional steps or tools after merging the clips to keep everything consistent?
Thanks again for your help
@yamypro Example for 15 sec: deactivate the Rife node at the top left; this one is designed only to interpolate the first 5 seconds of video, so for more than 5 sec, it is useless. Then, activate the 2nd segment and the 5th one on the Fast Group Bypasser node. This results in: 1st segment (5 sec), 2nd segment (+5 sec), and 5th segment (+5 sec), totaling 15 sec. On the 1st segment, you have deactivated interpolation, so your video will be 16 fps. The 2nd segment will also be 16 fps, as will the last one. However, the 5th segment is used to join the videos into 15 sec and then interpolate to 60 fps (you can choose the number). Keep in mind that it is only in the last segment that the videos are merged and the 60 fps interpolation happens in the workflow. So, logically, for a 10-second video, just activate the 5th segment (the 1st one is always activated anyway).
Hey! Fist off, thank you for your work!
I am having trouble understanding how to generate extended videos beyond 5 seconds.
For a normal video I only have "enable 60 fps" in the fast bypasser node enabled. I figure I need to "enable 2" and "enable 5 rifle" for two videos to get merged together? If so, do I need to add a new prompt in the new section or can I also leave it blank?
Also it seems that, for some reason, after generating and interpolating the first 5 seconds, comfyui just seems to crash (not the UI, just the terminal)
I am really scratching my head here what I might be doing wrong.
Hello, sorry for the late response, I hadn't seen your message. Try the new version; it is simpler.
Super Power Lora published to Comfy manager. It seems even nicer than the one in the pull request for RGthree. Should be able to use it for swap out of the one in the workflow. https://github.com/HenkDz/nd-super-nodes
Absolutely love this workflow. I have been playing around with this. Can you provide some examples of videos which you generated with multiple 5s videos. At times, I am facing difficulty in maintaining the face from the image in the 2nd/3rd videos. Can you suggest how best we can chain the prompts to maintain scene consistency
Hello, I have a few examples, but they weren't made using only my workflow; I used several third-party tools. It’s not possible to do it with just my workflow, but let’s just say the various stages were completed using it. https://civitai.com/images/114733916