Hey honey, what are you doing...? 😮
... 😨
Wanna support me?
Don't send Buzz, send me USDc or USDT through Polygon Network:
0x3f1e1323fa90b4c9f5a386bfa5b1e72f5a6c25e9$0.10 or $10, any amount is welcome and helps me make more LoRAs for the community
Workflow
I don't use ComfyUI, so I don't have a workflow to provide 🤷🏻♀️
The training is done on Wan 2.2 A14B Diffusers (BF16)
Training
320 steps on high noise
2 edited videos (80 frames each video)
HLR + LCD (These acronyms were created solely to confuse you)
Lots of confidence
$5 (✋🏻😐🤚🏻)
Usage (Low noise)
This LoRA was trained only on High Noise, which means you'll have to use some Low Noise that knows what a penis is (I used the Low Noise from DR34ML4Y V2).
Got an idea for LoRA? Then let's do it! (or at least try 🤠)
In my repository, I created a model called "LoRA Requests", so if you want to request a LoRA, send videos and photos to help build the dataset for a public LoRA that will be posted here, that's the place to go.
Open LoRA Requests on HuggingFace 🤗
Uncensored prompts and examples (cut to action)
To see all "cut to action" prompts, visit this LoRA on HuggingFace 🤗
I2V:
Prompt (ongoing): You can combine or use one of these
He holds his hard penis, moving his hand up and down stroking repeatedly.
---
She then touches his shoulder, and he looks back worried and scarred.
---
She holds his hard penis with her closed fist, moving up and down stroking repeatedly.
---
She leans forward and holds his hard penis with her closed fist, moving up and down stroking repeatedly.Prompt (cut to catching masturbating):
No man, not here ✋🏻🤠Prompt (cut to assisting):
No man, not here ✋🏻🤠Prompt (cut to catch and assist):
No man, not here ✋🏻🤠High Noise LoRA Scale: 1.0
Low Noise LoRA Scale: 1.0
Shift: 4
T2V:
Technically it works, but I haven't tested it. If you're going to test it, use a low scale, starting at 0.5 on High Noise LoRA. Use the structure of I2V
About HLR + LCD
This is a fast-learning technique, which makes LoRA less flexible. This can drastically reduce creativity, but it yields stable results while using few resources. Negative effects:
It will probably only work at the same camera angle
High chance of not responding to different prompts
High chance of forcing the original characters of the training video Training only the High Noise reduces this chance, but it can force things that are VERY similar, like the same hair length, same color, etc.
Download
Download High Noise LoRA (From HuggingFace)
Download Low Noise LoRA (DR34ML4Y_I2V_14B_LOW_V2)
Help me creating more
If you want to help me continue making LoRAs, or if you want me to make a LoRA for you, buy 5000 PlayCoins at Aisha-AI and transfer them to my account (account number 2).
This helps Aisha-AI to stay alive and produce new LoRAs for you all 💜
Description
First version
FAQ
Comments (13)
you always do great work, just a feedback on this is that it massively change the character face! keep going mate :-)
If it's switching, then there's a problem with your workflow. I do several tests before posting, which is why the examples always show a white and an Asian woman. If it's switching, then there's a problem with your workflow. I run several tests before posting, that's why the examples show a pale Asian woman, an African woman, and a redhead: consistency of appearance test.
Please use the standard WAN 2.2 workflow.
None of his Loras work for the blink. Faces look completely different. He just claims it works and we are doing something wrong, yet there is not a single person having success with this.
@AI4L1VE I'll record my screen and post on examples to show you that it's simple works. Just wait for it, and take a look on other's posts also
@AI4L1VE https://civitai.com/posts/27089160
face is not an issue, it all works fine if you know what u are doing with your wf, consistency is very good, the real problem is anatomically incorrect penis with whatever LoRa u use at low or try to add to High to fix it, Frenulum, a part of penis is supposed to face the viewer and be opposite of the man
@erdelman73267 I don't think it needs to be perfect. This should be done to fuel a fantasy, not to imitate reality. I make "cut to" versions so people can fantasize about other people. But creating hyper-realistic videos down to the smallest detail starts to get dangerous.
I'm not saying I do it on purpose... It's just that I don't see it as a really big problem.
Have you used it with Lightning LoRA?
@g1263495582 I don't
@AishaAI That's exactly why you don't run into the face-changing problem
@g1263495582 Perhaps that's why I keep repeating to use the standard WAN 2.2 workflow, which OBVIOUSLY doesn't include any other LoRa 🤷🏻♀️
@AI4L1VE
High sampling:
- cfg : 0.8
- euler_ancestral
|-- beta
- steps : 6
|-- start 0 - end 3
- this lora : 1.00
Low sampling:
- cfg : 0.8
- euler_ancestral
|-- beta
- steps : 6
|-- start 3 - end 6
- DR34AML4Y v2 lora : 0.9
Done!! 😌
@AishaAI Nah, the standard ComfyUI workflow already comes with Lightning LoRA attached by default though 🤔 And sure, use a standard workflow without Lightning LoRA — but you never said anything about not using a model that has a merged LoRA with Lightning LoRA baked into it too 🤫.