loaded the work flow from your example videos to se how to use the models you recommend, but its loading as a generic wan2.2 workflow, do you have one with the recomendated models loaded?

Your suggested model file (https://huggingface.co/lightx2v/Wan2.2-Lightning/blob/main/Wan2.2-T2V-A14B-4steps-250928-dyno/Wan2.2-T2V-A14B-4steps-250928-dyno-high-lightx2v.safetensors instead of wan 2.2 high) does not work for I2V as it is T2V.

What is the function of this lora?

https://huggingface.co/QuantStack/Wan2.2_T2V_A14B_4steps_25-09-28_Dyno_High_lightx2v-GGUF

Edit:[Here is the GGUF versions of your suggestions. I found them on accident after looking all over so I came back to share as this is the only way I can do most things and I'm guessing many others are in the same boat.]

Hey this is pretty impressive. I have tried these other universal deals and still they want the trigger words which I have grown to despise, but this actually had my images doing naughty things with no lora and natural language and that is tight so kudos. What even brought me here was seeing some videos in the feed out there that were so unique that I was asking how did they pull that weirdness off and your lora was all they used. =P

Absolutely amazing, your lora does what most lora’s can’t, provide the intended action without wildly morphing up the female in the prompt.

How do you get good looking penises in Wan version? It's generating mangled, deformed penises even if there is one in image.

Would u mind to tell me that how u caption ur data?Im trying to make a lora like yours for learning purposes.

Why are nipples added to the buttocks in the WAN 2.2 model? I've tried to fix it, but I can't find a solution.

A bit of a odd question but can I make women lactate and squirt breast milk with this flux and Qwen checkpoint? Or do I need a different Lora.

I get oversaturated images, why?

I tried with the example image workflow, I installed the missing nodes and the result is a total let-down, the image is extremely blurry, you cannot distinguish anything... Anyone got some tips on how to use this? P.S. I am using flux1d

This doesn't work on qwen, image doesnt show up, only noise

这是一个FLUX，wan，qwen都能用的lora？

What checkpoint do you recommend to use? With your images, do you use just a default checkpoint/model

The LORA works fine, but often the nipples show through the clothes — what could be the reason?
I’ve already tried several settings.
Example IMG

Your workflow is not working. Like others have reported, I get an extremely blurry/hazy output with next to zero, and I do mean nearly zero, movement. Person just kind of slightly turns left and right.

I used the workflow attached to the 2nd image, the one with the girl licking her own boob. Kept all settings at default on a RTX 4090 with 64 GB RAM. I've tried more than one image, too.

Any idea what might be up that can be fixed?

I’m curious to know what it will mean to be a result on my femboy’s

This model is excellent for my first attempt on transgender people with chastity cage A big thank you alcaitiff

Is the strength recommendation correct? 1 for both HIGH/LOW?

where do u have the workflow for your wan 2.2 i2v model i cant figure out how to apply it. i have the default comfyui template with loras between mdoel and sd3 and no nudity is rendered

interestingly it does not work with qwen edit 2509, BUT it does work with the nunchaku version of qwen edit 2509. Might have something to do with the custom lora loaders for nunchaku

I hope to see this Lora for Z image , I saw great images with it on wan, krea or qwen :)

How I got it to work:

Comfyui, then download the "high" and the "low" from this website. Then put them in between the main model and the loraupdatemodel in a "load Lora" node. also connect the clip to it. Voila it works.

What is the trigger string for Lora?

Z-Image turbo needs this. BTW your recommendation of Ostris AI Toolkit really helped me. I finally train a local LoRA for Z-Image

Right click save videos, images as ... no workflow found in comfyui?

I think I already know the answer for this but I have to ask anyways: are you considering a Flux2 version of the Mystic lora? I know it's a beast to train anything on in the first place, was just curious if you had looked into it at all.

Flux v7 does its nsfw job, but i feel like a lot of generated images, have same body and face structures, doesnt apply other types of variations for some reason.

Can I find a ready-made workflow for i2v somewhere?

is it working with Qwen2512 T2I? It doesnt work for me, it just gives me a big blob of artefacts

One of the best I’ve seen so far. Can’t wait to use it. To save up some buzz.

You think you could retrain a new T2V version where faces have been masked out of the dataset? If not do you mind sharing the dataset?

As normal lora it doesn't work (i2v). There is also no workflow, I couldn't find it.

KLEIN 9B please sir

Which workflow do you use to work with this??

Please specify the version of Qwen this is for.

who can explain how to do a good workflow for this

Can we get any information on size of data set and training steps?

I would like to understand the training parameters used in the MysticXXX v3 model, including details such as dataset size, image quality, and other key training configurations. I am working on a similar use case, but I am currently facing issues like anatomical inaccuracies, body distortion, and the generation of extreme NSFW content during image inference after fine-tuning.