⚡ 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 (126)
Not working correctly for me, i get dozens of lora key not loaded messages (for SVI), so all my tests give me garbage output. Also the "second video" node keeps activated, when i disable video 2,3 and 4.
Just use the SVI lora model from Kijai repo in hugginface. I had the same problem. Here is the link: https://huggingface.co/Kijai/WanVideo_comfy/tree/main/LoRAs/Stable-Video-Infinity/v2.0
I installed every missing note but I keep having problems with WanImageToVideoSVIPRO node and FindPerfectResolution node. It says that the are missing and anytime I install them it keeps telling me that they are not installed.Does anyone have solution to this?
Same issue for me
Try to summon them from double click, if that doesn't work check terminal if they are imported at all.
Yeah, I am also having the same issue.
@HSDHC can't find them there. As they are not installed at all.
So I found the solution. First we go to custom nodes folder. Then we delete the kjnodes folder. Then we clone from the github repo here is the command - git clone https://github.com/kijai/ComfyUI-KJNodes.git
For FindPerfectResolution we clone - git clone https://github.com/ashtar1984/comfyui-find-perfect-resolution.git
That should be it. Have not generated video yet. At least the missing nodes are gone... for now. Please comment if this has worked for you.
I had the same problem. Delete the ComfyUI KJNodes folder. Download the zip from https://github.com/kijai/ComfyUI-KJNodes.git and unzip it into the custom folder again. this worked for me.
@noreta587 It worked for me, but now I'm getting this error:
Prompt outputs failed validation: PathchSageAttentionKJ: - Required input is missing: model PathchSageAttentionKJ: - Required input is missing: model
@lolbleach001584 Sorry, I disabled SageAttention and added a note on how to enable it. Please redownload it.
Prompt outputs failed validation: PathchSageAttentionKJ: - Required input is missing: model PathchSageAttentionKJ: - Required input is missing: model
Sorry, I disabled SageAttention and added a note on how to enable it. Please redownload it.
@taek75799 Hi, I have SageAttention installed but I'm still getting the same error.
@lolbleach001584 I disabled it, try without it. If it works, check the subgraph where you select the SVI LoRAs; it's located in there. Perhaps the nodes I used aren't working for you, I'm not sure. Try replacing them with others that work.
@taek75799 Hi, thanks for replying. I get this error when I deactivate it.
Circular reference detected while resolving output 0 of node 188 (Get_vae). This creates an infinite loop in link resolution. UniqueID: [undefined:188[O]0]
@lolbleach001584 It seems that the error is coming from "Get VAE". Try deleting it and connecting the "Load VAE" node directly.
Sorry to those who had issues with SageAttention. I've updated the workflow and added a note on how to activate it if it is installed on your system.
So I'm getting an issue with the transitions, I activate Section 1 and Section 4, the 2 sections do a transition but with a crossfade blur. They don't stay consistent across the 2 videos. I've kept workflow as default, all apart from using your GGUF Q8 model. Prompt is consistent across the 2 sections. No errors show in log, I don't know if I've missed something or if there is a trigger word I've missed.. Thanks for all your hard work @taek75799 appreciated.
Try the transition with the 1st and 2nd sections. Unlike my old workflow, there's no need to go through the first and last frames; by doing the 1st and 2nd, it should join the video using the "VideoHelperSuite" node (the very last node, not section 4), hopefully it works.
By the way, thanks for your feedback. I haven't tried the 1st with the 3rd or 4th yet; that might not work.
just tried it, and indeed I have a cut between the two scenes. So it needs to be done in the right order.
@taek75799 awesome! I thought that might have been the case. After I posted this, I went back in and re-read your instructions in the workflow on the far left stating 1st and then 2nd sections to be done in order. I didn't have a chance to try it out before leaving the house. Glad you got back to me though and clarified it. Once again I appreciate the work ;)
hey there, I tried to open it (just downloaded it) and I keep getting Wanlmage ToVideoSVIPro in subgraph 'SECOND PROMPT Subgraph', updating from the manager does not seem to help either. I also tried deleting the custom folders and install these via git without any luck https://github.com/kijai/ComfyUI-KJNodes. https://github.com/ashtar1984/comfyui-find-perfect-resolution.
It's annoying, many people are having issues with installing certain updates. So, if I understand correctly, the problem is linked to a "WanImageToVideoSVIPro" node? Do all the other KJNodes work? If so, I really think it's an update issue. Try updating your ComfyUI as well, not through the manager, but using the update folder within ComfyUI itself.
@taek75799 thank you, updating desktop comfyui worked, I'll try the workflow soon x
I'm sorry but I don't understand what I'm doing wrong. I've set the WAN 2.2 Enhanced NSFW in both high and low loaders and the output looks very bad. Loads of noise and poor quality. What am I doing wrong?
Hello, can you send me the workflow with the settings you use? Often, people use my workflow, but the settings are not correct. https://www.swisstransfer.com/fr-fr
Just wanted to let you know that princepainter uploaded a new PainterI2Vadvanced node. I haven’t tested it myself yet, but it could be a good addition to your workflow.
Dual-output design: simultaneously outputs two sets of high and low conditions, perfectly adapting to the dual sampler architecture
Intelligent Color Protection: Prevents greenish and dark issues common at high dynamic amplitudes
Zero Dynamic Loss: Post-processing correction strategies that fully preserve the original dynamic enhancement
Adjustable Parameters: Finely control dynamic amplitude and color correction intensity
Intuitive Naming: Output ports are clearly labeled for easy workflow connectivity
Hello and thank you very much, I will check.
Hey, thks for your work, what happend with your svi workflow, is it was removed?
Hello, yes I deleted it, there were still too many issues with the nodes. I should be updating the V2 version of the workflow fairly soon there has been an update to PainterAdvanced.
@taek75799 uh, will wait
@taek75799 It works out great after I did some tweaking of it all. The WF "bones" worked, but I cleaned up some of the nodes. I also added a sharpened that can be toggled.
Thanks for the great WF and I look forward to v2
@MankyPoodle Hello and thanks for your feedback. Which workflow are you talking about? The Lightning Lora Video V2 which is still here?
@taek75799 sorry - the SVI Pro one you releases yesterday
@MankyPoodle Well done! Are you using a GGUF?
@taek75799 nope, using the FP8 version of your NSFW checkpoint
@MankyPoodle Thanks for your feedback.
Is this following this implementation? Have you seen this? https://www.reddit.com/r/StableDiffusion/comments/1pzj0un/continuous_video_with_wan_finally_works/
Thanks for the I2V workflow and the model. How can I modify this workflow to use T2V with this workflow? Would you recommend that, or should I got for some T2I workflow first and then use the output in this workflow?
Hello, my workflow is based on nodes that only work with I2V. You cannot transform it into T2V. You could try this: put a black image into "load image" and add a prompt; this should turn the model into T2V. I have never tried it myself, but I have often seen this workaround.
@taek75799 Yes, using a black image worked for me.
@taek75799 I am trying to get specific facial structure (ethnicity). But most such loras are available for Flux and SD models. How would you recommend achieving that with WAN I2V/T2V models?
@rahulr1199 I am not very familiar with T2V. For I2V, I use either Pony, Illustrious, or Chroma for NSFW for different ethnicities. I think the LoRAs you mentioned would be ideal. Good luck to you and thank you for your comment.
I'm just trying to understand something: I don't see the two SVI LORAs actually being used anywhere in the workflow...it works, but am just wondering if they aren't there? Or I'm just not seeing them?
Hello, they are in the subgraph at the top left. In principle, you should be able to see them in order to select them.
@taek75799 Hmmm, this is the Lightning Edition V2 workflo?. I only see three subgraphs, the three for each video. No loras being loaded there, and the three power loaders on the main graph don't have the SVI Loras. Which "top left" subgraph do you mean?
@Bojo66 he workflow you are mentioning isn't the one with SVI. I deleted it because many people were having trouble installing the node required to run SVI. Another issue: it works very poorly with certain merges, and above all what I consider the biggest problem is that most of the videos are in slow motion.
I want to create a long video, so after creating the first video, I activated the second button, wrote the following in the second prompt box, and created it, but the video comes out with the same starting point as the first video. What should I do?
Hello, which workflow? The Lightning Edition V2?
@taek75799 Yes, that's right
@asfg2 Hum, in principle, the second video will be the first and the second ones joined together. Try regenerating while leaving the first two sections enabled to see.
@taek75799 I was mistaken. The second video was already merged, but I ran the prompt in a different pose, but that video was repeating the same movement, so I didn't notice. I'll try changing the prompt and try again. Thank you.
i just wrote another comment with the same problem, in my case the second video is not merged, just the same 5 secs video.
Ok i can generate the first 5 secs video with no problem even at 60 fps, but when i check to generate the second video, even with different seed and prompt it just generates the exact same video that was in the first part. Any ideas?
Hello, can you send me the workflow where the problem occurs? I would just need to add an image and a prompt, then run it to see if I get the same result: https://www.swisstransfer.com/fr-fr
Check each video in your output folder carefully: there should normally be one 5-second video and another 10-second video.
I am using v2 with every setting at default, the only change i made is linking the gguf model nodes you provided instead the safetensor ones and there is no 10 seconds video, even the video is exactly the same in the preview inside the workflow.
I think it does not try to generate the second video because it takes mere five seconds to create the second video, like it is copypasting the first video.
i need some explaining how does this work, i cant be onlyone who dont get it. Do i need to enable anything at all when making the first video or enable second 10 sec video first since not recommended to start with 60 fps first video. Really confusing in the note.
Hello, there's nothing you need to do to start: for the first video, just launch the generation. Then, if you want to continue it for an additional 5-second extension, just activate the second section, and do the same for the third section if you like the first 10 seconds (5 + 5 sec). Now, when I say it's not necessary to enable 60 FPS for the first video, I mean that if you don't like the video, there's no point in wasting time on 60 FPS interpolation. If you do like it and only want 5 seconds, then enable it. It's just a tip to save time.
@taek75799 What happened to me is: i made a video without enabling any of the groups, got the video, made it 60 fps using the "enable first video 60 fps", got the video, disabled the "enable first video 60 fps" then enabling "second video 10 sec" and wrote to the second prompt what i wanted to see and something horrible came out of the video. There is seed high and low and they all have number 0 i didnt touch any of those.
A video guide would be appreciated with this kind of workflow for people like me
@Halmes Was the quality really degraded compared to the first video? Try increasing the resolution and, if you can, upscale your image. I will post an update this Saturday.
@taek75799 the second video was just different from the first, or it continued but in horrific way.
I had great results with your SVI workflow, using your GGUF model. But I lost the workflow when I accidentally nuked the wrong drive when installing linux. If you could please send it or re-upload it as a beta test or something, I would greatly appreciate it!
Hello, I sent you a DM. It seems that SVI works very well, but only in 480p. To be honest, what really bothers me is the slow motion of the LoRA; my models were specifically made to eliminate this effect, and I feel like I'm going back in time with this LoRA.
I'm having trouble with a missing note and comfyUI won't tell me how to get it. Did anyone else have that problem?
FindPerfectResolution
PainterLongVideoin subgraph '2 Subgraph'
PainterI2Vin subgraph '1 Subgraph'
When i click on Install All, ComfyUI acts like it has been installed but when I restart it, I have to install it again... and again... and again.
I can't be the only one with this issue, can I?
Try to install them manually: https://github.com/princepainter
https://github.com/ashtar1984/comfyui-find-perfect-resolution
Hello my problem is created video doesnt follow my instructions like i say insert but it doesnt insert it ooly shows the genitalia
Hey taek, thank you for the workflow. It works nicely with your Enhanced NSFW model. But I'd like to have the option of using a different model without a baked in speed lora, just for flexibility.
Do you have any idea where to put the speed lora loaders in your workflow? Putting them before the subgraph for the SVI loader works but the video is still a bit blurry afterwards.
Do you have any idea?
Hello, you need to place them in the first subgraph where the SVI LoRAs are. If you do it this way, those LoRAs will be active for every section. You just need to add the same "Load LoRA Model Only" node in the subgraph, I believe. Regarding the video blur, I'm not sure; SVI still doesn't work well with merged models, so that might be why. I haven't had enough time to test it properly yet.
@taek75799 Thank you. I've tried that and it doesn't work with models without baked in speed loras. It's like the speed loras are ignored, even though they're loaded correctly before the SVI Lora.
Without speed loras your Enhanced NSFW Q8 GGUF works flawlessly though, SmoothMix for example works too.
It's only the "standard" models which need a speed loras which give the blurred output. But anyway, the workflow is awesome and with your model it works great, as I've posted below.
@fatberg_slim Thanks for your feedback, your video is really good. I'll try to take a look at it when I can.
taek, where can i find "steps" ? did u rename it to something else or.. help pls tysm
Hello, on which workflow?
@taek75799 this one taek
wan 2.2 i2v
have the same problem, don't know if its node conflict that changes it
@miikiious375 hi, Go to the subgraphs (the nodes with the prompt). There’s an icon at the top right of these nodes; click on it to open the subgraph. You'll find the KSamplers with the step settings inside
@taek75799 Thank you very much, haven't thought of that, found them :)
I feel like an idiot, I have messed with the seed node, clicked Randomize Each Time, or New Fixed Random but I always get almost the exact same output when using the same lora and prompt. I am trying to run it again with a different seed to see if i can get a better video but always the same output. Is there a way I am missing how to tell it to get a new seed?
In principle, Wan does something different if the seed isn't the same. If the video really stays exactly the same, it's because the seed remained identical. Set it to "randomize in time" and, in principle, you'll get -1.
@taek75799 Thanks for responding to all our silly questions, you are awesome! Ill try that again but it felt like it was identical each time. Felt crazy.
What about just generally pixelated look to the output? 60 fps smooths it out but in general the resolution doesnt feel very high even if I use a pretty high res starting image. I am reading other comments and trying to figure out what setting that may be to get a more high res output. Tried a couple different input images and they are all on the pixelated side.
@bm939 The workflow is set to 720 by default, but I do my previews in 1024. If your GPU can handle it, increase the resolution and go with the Q8; it's the best version.
@taek75799 ah ok that makes sense, i have a 5070ti so i will try. HOPEFULLY last question, still struggling with the seed thing where output is the same but ill keep messing with that. The other issue is enabling the second video, it starts at the same start point of the reference image which i think is intentional? Not sure? But the problem is that it generates pretty much the same image as the first video and doesnt really follow the second prompt, but whats way worse is a few seconds in the whole video just pixelates and blurs out into some horror show thing lol. Any ideas on that?
@taek75799 ok i think i get it more, so the second video generated actually has the first 5 secs as the first video, so what it seems like is when the true second video starts, its turns into a blurry pixel horror thing and I dont know why lol
Does anyone know if the SVI PRO SIMPLE workflow was taken down or if the author deleted it? Also, does anyone happen to have a backup?
This is a GREAT new workflow! I'm still testing but first tests are blowing me away! You are a god! :)
I'm just learning to use ComfyUI on my desktop. I'm trying to make this workflow run, but I get the error "Prompt execution failed. Cannot execute because a node is missing the class_type property.: Node ID '#170'"
...and I can't find a node with number 170. Other's numbers I can see, but FindPerfectResolution doesn't appear to have one. Also, when I open this workflow, Comfy says every time that the missing nodes were installed. I tried deleting and redownloading the workflow, checked that I have all nodes installed via manager, checked for updates etc. Comfy should be up to date. I'm using WAN 2.2 Enhanced NSFW Q6K and FP8 version of the recommended NSFW Clip. What am I missing?
And now I saw a notification "Workflow contains unsupported nodes" and FindPerfectResolution is highlighted in red.
@jerkoffalltrades Hello, try installing it manually using: git clone https://github.com/ashtar1984/comfyui-find-perfect-resolution
@taek75799 Hello, I had the same issue. After manuallyinstalling the FindPerfectResolution node, I have the error "Cannot execute because a node is missing the class_type property.: Node ID '#298:287'" This is an error in the main subgraph where several sampler nodes are not connected after the subgraph is expanded (unpacked).
Prompt outputs failed validation:
KSamplerAdvanced: - Required input is missing: negative - Required input is missing: positive KSamplerAdvanced: - Required input is missing: negative - Required input is missing: latent_image - Required input is missing: positive
@reloadUI Try to update ComfyUI, not with the Manager, but in your ComfyUI/update folder.
@taek75799 no updates found
@reloadUI Remove FindPerfectResolution and set a resolution manually in the Painter node to see if that is where the problem is coming from.
EDIT: I manually cloned the repos:
git clone https://github.com/princepainter/ComfyUI-PainterI2V.gi
git clone https://github.com/princepainter/ComfyUI-PainterLongVideo
Now the workflow begins, but it stops midway with an error:
KSamplerAdvanced
The size of tensor a (48) must match the size of tensor b (16) at non-singleton dimension 1
below issue seems solved:
@taek75799 I was going to try your suggestion. Previously, I had shut down comfyui so I had to start it back up again. When I did and re-opened the workflow (that I had previously saved before shutting down), I saw this message:
This workflow uses custom nodes you haven't installed yet.
PainterLongVideoin subgraph '2 Subgraph'
PainterI2Vin subgraph '1 Subgraph'
So I reinstalled them hit apply, re-saved the workflow and shutdown/open up comfyui. Once AGAIN, I dee the same message above. I think this is likely the issue. Any idea why these nodes appear to be not installing properly or uninstalling? What could be happening here? Thanks again.
@reloadUI I've already had some issues with certain nodes; installing them manually fixed the problem. Try unpacking the subgraphs: right-click on the subgraphs and select 'Unpack Subgraph', as there are still some problems with them.
@taek75799 I am running it but it stops early, with the error:
"KSamplerAdvanced
The size of tensor a (48) must match the size of tensor b (16) at non-singleton dimension 1"
I am using :
wan22EnhancedNSFWCameraPrompt_nsfwFASTMOVEV2FP8H
wan22EnhancedNSFWCameraPrompt_nsfwFASTMOVEV2FP8L
nsfw_wan_umt5-xxl_fp8_scaled
wan2.2_vae
afaik the subgraph looks ok, but it is where the issue is coming from
@reloadUI Use the Wan 2.1 VAE and it should work.
Manually installing FindPerfectResolution, PainterI2V and PainterLongVideo solved my situation. Cheers!
Where would the download be for the 2 safetensors?
wan2.2\Wan2_2-12V-A14B-HIGH_FP8_(lightning edition)_V1.1.safetensors
wan2.2\Wan2_2-12V-A14B-LOW_FP8_(lightning edition)_V1.1.safetensors
You may want to use the High and Low .gguf files from this Workflow's author instead of the safetensors: https://civitai.com/models/2053259
Put them in your ComfyUI\models\unet folder then use the Unet Loader (GGUF) HIGH and Unet Loader (GGUF) LOW nodes in the workflow to select them (you may need to restart ComfyUI for the .gguf files to show up). Disconnect the Load Diffusion Model HIGH and Load Diffusion Model LOW nodes in the workflow and connect the Unet Loader (GGUF) HIGH and Unet Loader (GGUF) LOW nodes in their place.
On second thought, what I said may be wrong... I haven't been able to get these workflows up and going. Not sure what I'm missing in the workflow.
@CiRender i have yet to get anything in this submission to work either.
For those that have bad results for the second vid, make sure to not include loras from the first video since they carry over(i assume? im also a newbie lol.) At least for my case when i untoggled the repeated loras, or simply not adding loras for the second video works for me. Hope this helps.
There was an error in the Lightning Edition V2 workflow: the LoRAs from the first video were also being applied to the subsequent videos, which was impacting the quality. Sorry about that.
OK, so now I ran into a new problem. When I load LoRAs, they load normally, but the box in the node is stuck behind the "+ Add Lora" button and I can't change the weight or anything. Just add new Loras.
Hello, that's annoying. I've never encountered this bug. You can try replacing the 'Power Lora' node with 'Load Lora (Model only)'; it will work the same way and you can stack them.
Replacing it with LoadLoraModelOnly did work. How do I stack them? Just make a chain of high and low loras before the model input connection in the next node?
FYI: adding new power lora nodes produces similar nodes, where all I'm able to click is the + add lora -button. Also, the fast group bypasser nodes has somewhat similar situation. It shows four stacked fields, and I can only use the top one, which says "enable second 10sec video". This might have something to do with rgthree node installation success then?
@jerkoffalltrades Without Nodes 2.0? I mean don't use Nodes 2.0, rgthree does not support it
@qek Thanks! Disabling Nodes 2.0 helped.
"This version is missing files"
Thank you for pointing that out, I had to delete it to fix an issue related to the first LoRA.
There was an error in the Lightning Edition V2 workflow the LoRAs from the first video were also being applied to the subsequent videos, which was impacting the quality. Sorry about that.
In this updated version, the First Lora High and Low nodes are not connected anywhere. And how should I approach the Set_MODEL -nodes, which seem like a dead end? Can I just leave them as-is or should I make an output connection for them somewhere?
Nevermind. I just found the hidden nodes behind the "new subgraph" node.
Lightning Edition V2 this applies to the workflow wan2.2_LIGHTNING-EDITION_long-video_FP8GGUF.json ??
Can you tell me if negative hints work in the wan2.2_LIGHTNING-EDITION_long-video_FP8GGUF workflow?
@cantayou527 It will work for any Wan model, but there are certain conditions to follow, and keep in mind that it will double your generation time. Look here in the description to see how to do it https://civitai.com/models/2053259/wan-22-enhanced-nsfw-or-camera-prompt-adherence-lightning-edition-i2v-and-t2v-fp8-gguf
Love this workflow. Just wondering if you had a workflow that extended beyond 15 seconds. Also character loras seem to have a bad effect - is there a recommended weighting for those?
I'm running this on a L4 with 32gb ram, and it is crashing on the third video generation ):
Any idea why? Should I use the purge vram node between videos?
I don't know, try it, it might work.
I don't know if its just me but with the latest workflow the coherency between the different videos feel less consistent than the SVI workflow, also the combine seems less effective in hiding the seams between vids, not sure regarding the animation speed tho, still pretty nice
Indeed, SVI is very good at making transitions, but it has other issues, such as the fact that videos are often slowed down. I created this model specifically to avoid those slowdowns. Personally, I prefer making multiple videos and joining them with Fun-Vace: the transitions are perfect and there's no degradation from retrieving the last frame, but of course, it takes much longer to do.
Looks like we don't have an active mirror for this file right now.
CivArchive is a community-maintained index — we catalog mirrors that volunteers upload to HuggingFace, torrents, and other public hosts. Looks like no one has uploaded a copy of this file yet.
Some files do get recovered over time through contributions. If you're looking for this one, feel free to ask in Discord, or help preserve it if you have a copy.