⚡ 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 (57)
SVI LoRAs – Download Links
download error ((
Hello! I just tried and it works. Maybe there was a network error.
Trying the lightningedition long v2.1 but all videos come out static, literally generates like 1 frame (seems like last one) and that's it, any ideas why is that? Tried with loras, adjusted some settings, length, number of frames, no luck.
Same, my result is a single image.
Edit: Apparently for me the image forwardfing is not always working, if I connect the output image from first video to output directly it works.
How do I disable lightning loras in the 2/3 ksampler wf? I don't see a toggle switch for it.
Not able to find the "Wan2_2-I2V-A14B-HIGH_consistent_face_Q8.gguf" || "wan2.2\Wan2_2-I2V-A14B-LOW_consistent_face_Q8.gguf" anywhere on the web. You have them on the default WF. Source plz?
Hi. Did you find it or use standart Q8_0.gguf?
Hello, I am a beginner. May I know if I have downloaded it WAN 2.2 Enhanced NSFW | SVI | camera prompt adherence (Lightning Edition) I2V and T2V fp8 GGUF After the model was developed, High LoRa and Low LoRa were downloaded without any changes to the parameters. However, the generated video had a lot of noise, and I don't know why WAN 2.2 Enhanced NSFW | SVI | camera prompt adherence (Lightning Edition) I2V and T2V fp8 GGUF Where should it be, thank you for your answer!
This is probably the best Wan2.2 i2v SVI workflow on Civitai. Very clean and easy to use when understanding how it works. Just a question: what would be the best settings when disabling the lightning loras? 3 samplers or 2? I tried to make it work by also raising cfg to 4 and steps to 25 on samplers but without good results. What I am missing? I am using the base fp16 i2v model.
Hi. I guess you have a problem that only one frame is generated. To fix this just change the value to "1:-0" in all "image select" nodes
where is image select nodes?
n
@rkrmf1524161 you found them? i am looking at the workflow and i dont see those nodes
@rafurius2008104 you can find them in 1, 2, 3 and between 3 and 4 modules. By default they are collapsed
Impossible to find those nodes mate
@Hassan1er Don't worry. I did find them and the values they gave didn't work. I'm still experimenting and will update if I find a change. But this WF looks busted.
I figured it out they posted the values inverted it should be "0:1" without quotes.
Also they are found inside the 10, 15 second Subgraphs. Bottom left between the image_1 input for the subgraph and Upscale_Image_by node connections. The last one is between the 15 and 20 second subgraph sections. It sits between the output of the 15sec subgraph and an upscale image by node.
Alternatively and Optionally you can connect the "Index" pin on that "Select Images" node to the blank pin on the subgraph input list and then use that outside of the subgraph to set the value. If you do this just remember to clear the value on the "Select Images" node or it will override any values set externally. Again this is 100% optional and i only did it for trouble shooting.
Best workflow thanks alot
Super workflow, merci beaucoup!
I'd recommend adding one "get last frame" node after the "get upscal" node to allow users to continue/end their video cleanly using a FFLF workflow like painterlongvideo which allows video input for motion retention purposes (but isn't as good as SVI for long vids). I upscale the last frame to the original image and use it as start image, use the original image as end, and run FLF "backwards" to create a nice clean loop.
I'm running this on a 5070 ti with 16g of VRAM. After the first step with the K Sampler, comfy ui just says reconnecting and the process freezes. I'm not running out of VRAM so I'm not sure what's going on. Anyone have any advice?
Try increase the pagefile size and ask a chatgpt, gemini or grok how to set your "driver's not responding" tolerance to about 60-120 seconds from the default setting. Its not hard, and helped me a lot with reconnecting issues. If you doesnt run out from vram, my bet is the driver. Also try the stable studio driver for your card, and also forget overclock and undervolt if u use video gen.
I love the find perfect resolution node, but if a lora requires a specific resolution, how can we force the resolution to be something specific?
this is excellent. the triple sampler workflow worked flawlessly with very good prompt adherence. for a 'slow' scene i used 1.0 weights for both lightning loras. I also changed the negative prompts to match the suggestions in the original nsfw i2v.
Thank you for your work!
I downloaded the 2 or 3 sampler version. But i wonder why there are actually 12 samplers? Moreover, why the number of output for the gloabal lora setting is only 5? Where is the slow svi lora?
Thanks again for your patience!
Hi,
I'm trying to use your wf lightningedition long v2.1 with NSFW V2 Q6K H/L GGUF of yours. Unfortunately, when I do so, the outcome is just a single image/frame in poor quality. For now I'm just trying to create short 5s vid at least.
Can you help me fix that one, please?
Done?
Same problem here, have you fixed it?
@dewiant look at hardwire666 comments or demiart's comments, that should help fix this.
2or3 workflow - Do not increase shift
I am used to increasing shift for high model to 12 for some parts of videos that I generate. However, with triple KSampler, it makes the end product completely noisy and full of artifacts.
I even tried adding an extra path (ModelSamplingSD3, SVI Lora -> lightning high) so I could adjust the shift of the High Lightning lora model only, without affecting the nolightning high model (for that one, kept shift at 8). This fixed the noise, but lead to heavy color degradation within the first clip.
I have confirmed that the issue is with triple KSampler for me by adding the triple KSampler to my other SVI workflows. Looks like you just have to keep shift at 8.
Is there a way to reuse/repeat the previous seed in the newest workflow version?
OK, so I'll just select "New fixed random" for each generation if I want the option. Just talking to myself here...
System:
- RTX 5060 Ti 16GB VRAM, 96GB RAM
- ComfyUI 0.18.2 (Pinokio), PyTorch 2.7.0+cu128
- Running with --lowvram flag
Models I'm using:
- Checkpoints: wan22EnhancedNSFWSVICamera_nolightningSVICfQ8H.gguf and Q8L (from your CivitAI page)
- SVI LoRAs: SVI_Wan2.2-I2V-A14B_high/low_noise_lora_v2.0_pro.safetensors
- Lightning LoRAs tested: rank64 (1022) AND rank128 (lightx2v_I2V_14B_480p_cfg_step_distill_rank128_bf16)
- Text encoder: umt5_xxl_fp8_e4m3fn_scaled.safetensors
- VAE: Wan2.2_VAE.safetensors
Issue 1 - Lightning LoRAs don't load (both rank64 and rank128):
I get hundreds of "lora key not loaded" errors with PEFT-format keys like:
blocks.0.cross_attn.k.lora_A.default.weight
This happens with both GGUF Q8 and fp8 safetensors checkpoints. Note: the same Lightning LoRAs load fine (1053 patches) in my WanVideoWrapper workflow with Kijai's Wan2.2 Remix NSFW checkpoint.
Issue 2 - Without Lightning (2ksampler mode, switches set to False):
- With 6 total steps (3+3 split): I get only one frame (the source image) then the rest is blank/uniform color
- With 20 total steps (10+10): same blank result
- Lightning LoRAs set to strength 0.00, only SVI LoRAs active at 1.00
Issue 3 - With Lightning at strength 1.0 (3ksampler mode, True):
- Since LoRA keys don't load, the result is extremely blurry with seizure-like flickering
SVI LoRA warnings (I know these are normal per your other comment):
diffusion_model.blocks.X.cross_attn.k_img.lora_down.weight - these appear but you confirmed they're expected.
My questions:
1. Are the "nolightning SVI cf Q8" checkpoints the same as the "consistent_face_Q8" referenced in the workflow? Or are they different models?
2. Which exact version of ComfyUI are you using? Could this be a version compatibility issue with the LoRA loader?
3. Is there a way to make the workflow work without Lightning LoRAs (2ksampler mode) with higher steps?
4. Have you tested with GGUF checkpoints + Lightning LoRAs successfully?
Thanks for any help! The workflow concept is exactly what I need for my project.
https://drive.google.com/drive/folders/1RZfDFxQiazpJ0nS5Rzl15W03ErTj33xG?usp=sharing
in these 2 nodes how do i get these files?? i cant find them
Do I need to suck the information where what model has to be placed in ComfyUI out of my arsehole or is this information anywhere available?
LOL ComfyUI\models\diffusion_models
bro mad af
is it just me or the wf broken after latest comfy update? keep getting these -> AttributeError: 'NoneType' object has no attribute 'get_model_object'
If it looks like the workflow is only outputting one frame, or just a blank all black video. This is what fixed it for me.
inside the 10 and 15 second subgraphs. On the bottom left there is a node called "Select Images". There is also a 3rd Select Images node between the 15 and 20 sec subgraphs. It as at the bottom between the output of the 15sec subgraqph and an Upscale Image By node.
One of the values on these nodes is called "indexs". It will likely be a value of -1. This is the problem. It needs to be set to "0:-1".
Odds are the way the node handles arrays was changed causing the mismatched values to just do nothing.
This fix was posted in another comment but the values posted were inverted. As well it was a little vague about where to find the "Select Images" nodes. I just don't want take credit for someone else's work.
Hope this helps some poor soul.
Hi, I’m currently using a checkpoint that already includes Lightning.
In this case, should I use the High (Lightning) LoRA node or the High (non-Lightning) one for best results?
Hi, thank you for your work. Im using your model and workflow to create videos and was able to create one decent so I wanted to use your option to create longer videos but I don't know how to do it correctly. I don't understand when you say "Copy the second section (10-second video), including:
the subgraph and Get-set nodes
the LoRA High node
the LoRA Low node
Paste it next to section 2.
Connect the extented image output to the source image input 1 and output 2ksampler:false/3ksampler:true to the prev_ samples input of the new section.". Do you have a youtube tutorial where I can check how does that work visually? I copied the first "new subgraph" node with all the get nodes but I don't think that's how it works.
Been using this workflow for a while and it's awesome! I'm having an issue with making videos longer than 10 sec tho. If I enable the 3rd section (15s) onward everything immediately breaks when the video reaches the 3rd clip, random hands and people always immediately spring up everywhere. Nothing I've tried has worked to fix it. Does anyone else get similar issues or have I maybe missed something? I've tried different models as well but no success there.
cannot find this LOW or the High variant file anywhere
wan2.2\Wan2_2-I2V-A14B-LOW_SVI_consistent_face_nsfw_fp8.safetensors
me too
same here
I am so lost and confused on how to use this in ComfyUI. With the SVI workflow i'm missing eight models.
Lightning Edition long v2.1 doesn't work for me; it generates a video file that's only a few kilobytes in size, and the video is black (empty)
Okay so when I try to run it nothing happens, and I get the following error in the Command Prompt: invalid prompt: {'type': 'missing_node_type', 'message': "Node 'ID #961:947' has no class_type. The workflow may be corrupted or a custom node is missing.", 'details': "Node ID '#961:947'", 'extra_info': {'node_id': '961:947', 'class_type': None, 'node_title': None}}
Did the recent ComfyUI update just break something involving this workflow? Now instead of generating just the last stage, it rebuilds previous stages more often than not. Do I have to change some memory settings or something to get back to how things were? (= I can generate stage 1 and then move on to the trials and errors of the next stage, generating only stage 2 after 1 is finished.)
After struggling for days and troubleshooting with AI, my conclusion is that, at least for now, this workflow has various minor issues and author mistakes (or perhaps it broke due to recent updates?). I highly recommend using a different, simpler workflow instead.
@mizuaoi look at the demiart comments above yours for the fix.
@kirkout Thank you for all your help! Unfortunately, I still couldn't resolve the issue. The moment the image is converted into a video, the original image instantly turns washed-out/whitish, and the overall image quality degrades extremely poorly. I've tried everything, but this problem persists.
This thing is WAY to complicated... just want to make a simple good quality consistent face video... that's it.. I feel like everyone who makes these workflows know so much about all this that they don't really know how to dumb it down for those of us that get overwhelmed with all this.
Don't worry you get there. Prompt Deepseek or something else and ask simple instructions. That is how I got. Now I'm in the process of tweaking everything better and testing different workflows/nodes cause these models need tweaking for 5080 16 gb vram.
Unfortunately, it doesn't work for me (rtx 5070ti 16gb). Outputs only 1 frame after the first "New Subgraph". (((
There is a bug in the workflow where his "Select Images" node for Video 4 is outside of his Video 4 group, so it's not being properly bypassed, resulting in only the final frame making it to the combine step. Move that node and its corresponding "Upscale Image By" node into the Video 4 group and it should be properly bypassed.
I can confirm that this works, thank you @nivlet . I will have to add that I haven't gone the full 20 seconds yet, so once I do more testing, I can comment more. So far the quality is not that good, so I will play around with it. For reference, I have an Intel Core Ultra 9 275HX with 64GB ram, RTX 5080 (mobile which is comparable to a 5070 desktop), and I completed the 5 second video with interpolation to 64 FPS in about 7 1/2 minutes
how do you mean video 4 @nivlet ? like subgraph4 ? and I can't see select image 4 node either? I am pretty new to this tbh so if you can help, i'd appreciate it.
@kirkout we have the same system it seems
@senpai2233 +1 Would appreciate clarification on this.
Does anyone know a way to "plug in" a final FLF node? Its all thats missing from this WF