I tried using this on runnninghub but it gave me this error: # ComfyUI Error Report ## Error Details - Node ID: 226 - Node Type: VHS_VideoCombine - Exception Type: audit.RHAuditException - Exception Message: Porn ## Stack Trace ``` The RunningHub platform prohibits the following content: Porn Polity Abuse Illegal * Terror ``` ## Logs



can anyone please help me setup in runpod? I will pay you if someone could help me figure same thing in runpod and gets it running. dm me or replt mt comment where I can text you.

i got error using ur workflow any help ?

Prompt execution failed

Prompt outputs failed validation: ImageResize+: - Return type mismatch between linked nodes: image, received_type(PATH) mismatch input_type(IMAGE) MultiplePathsInput: - Return type mismatch between linked nodes: path_1, received_type(MASK) mismatch input_type(PATH)

awesome! love these models. any specific changes for 2.1?

Can´t wait for the Q8 Version :D

Great face consistency! Well done!

I pasted in the 2nd english template without addition and the QWEN node started to add double line breaks so the strings were skipping every 2nd node? I think it is because the template has line breaks in it when it shouldn't.

Also having a manual override to change a one or more prompt strings in the 提示词反推区 block directly before they are sent to the next step might be useful?

hello, can you tell me how the 2.1 version posted here differs from the 2.1 version posted by huggingface, the model's weight is higher there, and the second question is, does this version already have a built-in lora SVI2pro?

全自動反向的工作流程要怎麼關閉與跳過一開始圖片反推,由自己輸入Prompt

Hi can you help me ,I am getting this error, PainterI2V

module 'tensorflow' has no attribute 'Tensor'

First, your remix models absolutely rock! Your new 2.1 version does much better a face consistency, but has mutilated the male penis. I prompt a lot for a naked man to come into scene and describe the penis, and no matter what I type, it's a big blob. Ver 2.0 rocks it pretty good though. Switching back for now and hopefully you can update that.

I tested and the MoE ksampler seems to work fine with it too for simpler step balancing.

What's the possibility of replacing 1020 High noise with the 1030 version

For some reason, in most sex scenes, the man's penis doesn't work as it should, it often shrinks, swells, or behaves like a slug and doesn't want to crawl into a hole. :D

face changes a bit when using version 2.1

I get black screens for all frames in wan2gp for i2v. not sure why. used finetune file and both safetensor files.

After testing it for a bit the only thing I could ask for in the workflow is the option to hold at any step to tweak and run the that step again as needed instead of always running start to finish.

For example 2.1 is quite good at keeping faces but still doesn't sometimes but changing the seed, prompt or sampler can fix that. But if it fails on the 6th of 8 segments having to rerun from segment one to tweak something in segment 6 isn't ideal.

I can already do this manually but it's very messy and could be added straight into the workflow (i think). Maybe as a "set starting segment" option that holds the last made segments for every previous one or saving segments so they can be directly inputted into steps that don't need to be done again.

workflower always "LayerUtility: PurgeVRAM" error

I used 2.0 for weeks with great results. Best model I've found by far!

For some reason 2.1 has been much worse unfortunately, face consistency isn't as good, and it can NOT make a normal looking vagina. It creates gigantic clits and/or lips like monstrous looking things. No matter how I adjust the prompt, very simple or complex it does the same thing.

for me the only usable custom t2v checkpoint of wan 2.2, great work!

could be we get the fp16 version? (maybe i dont understand the model and t2v can only be fp8 in its usable form)

When using the Wan 2.1 I2V (Image-to-Video) model, how can I keep the original face consistent? Also, what trigger words should I add?

Version 2.0 I2V is the best! Version 2.1 I2V is terrible! :((

Edit: I fixed it!

(i2v) As the low model is reduced weight, I used Wan21_CauseVid_14B_t2v_lora_rank32 only in the low-noise at strenght 0.50 - 0.80 At the top of the LoRas I am using, and I am very impressed how fast the results are, when I had to wait before for good movement and quality around 40min, now I get the same results (512w896h) at just 33 frames in around 10min (or less) while before I always used 81-93-121 frames.

----- Previous comment -----
I think there is something wrong with the fp16 compatibility in the models, locally I tested it at "default" which is in a Mac M3 Ultra and it doesnt work correctly, in Mac i cannot use fp8. Later I tested fp8 in your online workflow it does work.

First of all: THANK YOU! I appreciate your hard work training such a unique NSFW model. (we sure need something like that for LTX-2).
Sadly skin always look weird if lifting shirt is part of it, there is no natural looking skin, always black dots or some weird looking skin. also the Breasts are NEVER natural looking, it seems like it's strongly trained on SILICON / FAKE boobs. I tried hard with positive and negative but it fails always.
I hope v2.2 will fix these 🙏

Do i need to use lightxv2 to get the 8 step generation or are they baked in?

I'd like to express my gratitude. This is the only custom model that produces excellent results without launching a rocket into space. The other models, no matter how hard I try, significantly alter the original image. It's a shame the adult models won't get an update. But many thanks for what's already out there.

why is my i2v output look nothing like my input image?

An awesome custom model (I2V)! Simple to use with very satisfactory results. 💪

After using 2.1 version for last 2 days. i have came to conclusion that 2.0 and it's gguf files are better than 2.1.

T2V 2.0 is include Lightx2v? the workflow not found Lightx2v? or i need add in lora?

Does anyone know why I'm getting green skin on the nude photo I upload for the I2V? How to fix that green-colored skin?

does anybody know why I'm getting added dress or towel when I upload a naked photo for the i2v?

I always get cumshot from the penis whenever I generate anything, how to fix this? I added negative prompts but it doesn't seem to work.

this keeps giving women balls 😭

With all those Wan 2.2 NSFW models my girls start fast humping motion almost from the beginning of the video even if they are alone in the frame. What am I doing wrong ?? No prompt seems to change it :(

how do yo load the checkpoints/diffusion models?? the template I2V wan2.2 doesn't see it.

Newbie (question). I have an RTX 5060 16gb. Can I change the models to FP8 versions? Or GGUF?  Or do they not make nsfw versions. Lol

I have another question. Using the model in SVI context I always see the model VERY eager to go back to the starting position. Like the girl undresses and then dresses again. Did you ever encounter anything like this? How can I prevent it? The model is AMAZING btw.

Wolklfower how should it be changed to the GGUF model?

All i get when trying to download this is 'Couldn't download network error' High version downloaded ok

Why are they humping by default? I prompted to stand still but the woman is humping the thin air...

I have the models from HF. What is the .x2Fc. in the filename here?

penises in my generations are missing the knob and it's quite frightening

First of all, I would like to express my sincere gratitude for the effort put into creating this model. However, I have noticed a few issues.

1. There’s a recurring issue where, during male mas***bati** and a c*mshot, c*m is mistakenly emitted from the character’s mouth as well. This was a severe issue in version 2.0.

2. In many cases, the model mistakenly recognizes other objects—such as fingernails or pens being held—as the glans of a p*n*s, causing c*m to be emitted from fingernails or pens instead of the actual p*n*s. This was a severe issue in version 2.0.**

3. There is also an issue where male c*mshots are not being properly implemented, unlike in version 2.0.

hi, im missing node Qwen3_VQA i cant found on comfyui and im try to search on google but not found please help

This is a F8 model that has not been scaled so F8 cores can't use it and default to F16. If you can, please release scaled models.

whats everyone use for a nsfw prompt generator?

Is there any way to get a basic glans/penis tip lick? Whenever I ask this version of WAN 2.2 for it, it just goes for straight blowjob.

high percentage in mutated genitals in this one compared to other checkpoints unless its already in the image.... nice adherence though in animation. Could use some more training

Next Lora .. seedance style Hollywood fight action

When I download the Wan2.2-Remix-I2V-Comfy-Qwen3 workflow, I don't see the Load Image Advanced node. I see the Coco Loader instead. Load Image Advanced is not an available node under Comfyui_Qwen2-VL-Instruct

https://i.postimg.cc/Nf6tf4hj/Screenshot-2026-03-12-at-3-04-10-AM.png

i like the high noise motion of this model but the overall quality of the video degrades so quickly into a blurry mess. Is there any way to maintain more sharpness with this model? It's really bad. Even if I uses a 1440p image to make a 1280p I2V video the end quality looks like 480p