V1 looks amazing thx!

Mmm, looks great!

So V1 doesn't need the assist file? And at what strength for V1?

Not sure what I'm going wrong, but this is constantly changing the backgrounds and or zooming no matter how I prompt.

Thanks for making this.

This is a great step in the right direction, and it is an improvement over base even at lower strengths like 0.4.

I tested v1 using close-up portraits I2v using the int8 pruned FLF2VA model at 0.4mp 15secs 8steps (er sde/ normal) with dareties turbo lora, where the Lora had to create the entire anatomy, I also only had this and the turbo Lora active. These are my findings:

- it needs greater than 0.7 strength to have a reliable chance to generate the clitoral hood, which then appears inside the vulvar vestibule where the urethra should be.

- At strengths greater than 0.5 it modifies the original subject, first physics bounce smaller breast are inflated to c cups (90% chance), and Hair length grows to waist level (60% but just base will do this also 10%).

- At strengths greater than 0.5 it also routinely ignored shot2 out of three , it just removed it every time. without the lora activated it 100% does all three shots so it is because of the lora.

- In extreme close-ups the texture and details that evolve upon physical interaction is perfect, wish you can capture it at medium full shot range .

V1 is stills only though?

this lora is for T2I or FLF or ref2VA ? cuz when i checked prompts there is no pic reference included in text field

any shot we can get a list of types included in the training? thx!

This lora is overfit, it add random noise and make output look dark, should reduce the strength. For now, it's the best h3 pussy lora.

Excellent work! Could you please advise LR and steps (epoch) for lora training? what king of captions did you use (detailed one, created by LLM, or shot one describing only pussy)? Thank you in advance

Hello, in the latest v1 version, is it possible to use it by simply loading this one 74MB lora file? I saw in your reply to others that version 1.0 uses strength 0.9. v0.5 has two files; loading these two lora files, one set to strength 1.0 and the other to strength 0.35, works very well.

Curious is you're using a third party service to make these loras or are doing it locally. I've been having trouble with training h3 locally with ai-toolkit.

Testing T2V: for a vagina LoRA, it seems to have a fairly strong face opinion, but the vagina results are decent. (v1)

Thanks for the LoRA, but I’ve found an issue. It doesn’t interact well with a-word. When I prompt for interaction with a-word, it ends up interacting with p-word instead.

If there's a man in same scene, he gets a vagina too. How to prevent it?

Stills needed if pussy is as reference image, and in motion in case there is no pussy visible in the first frame or reference?

I use v1 and weight 0.5. the pussy show but anus not show?

if the prompt use insert in anus. the movie result is insert in pussy....

How to idef anus and pussy by prompt?

Love it because it gets rid of protruding flesh out of vajay jay's (out of scope I know..lol), but man if you have a picture perfect one in i2v it completely destroys it. Can't turn down any lower than .15 or protrusions start again.

If possible since your lora is the only one that gets rid of them, isolate that cause and make a separate lora just for that?

When viewed from the rear, I'm getting vaginas that are front facing with clitoris on top. I'm trying to isolate the problem if it isn't your loras. I'm using both the still and motion loras as specified in the description.

Just to get it right, if I have an I2V situation where the vagina is already there, is the best practice to only use the motion lora at around 0.35 or V1 (img only) + the motion lora at around 1 and 0.35 respectively?