Looking forward to this one. Looks like a good first pass. Excited to try it!

i love you bro

Nice job! Please add a blink to front facing doggy too < 3

Bro, create a female masturbation Lora to beat him: https://civitai.com/models/2227772/suddenly-stroking?modelVersionId=2507936

His Lora's perspective isn't perfect; it's too close and too low, and the expression is too ferocious.

Works great but the annoying part is, if the character has clothes on or wears glasses they disappear. Any fix for this ? Also, maybe you can share your settings like lora strength, cfg, shift scale, etc. so we can get the best results. Maybe settings if we use accelerator lora. Thank you! Good job overall!

These recent loras work great! Any chance at an anal one?

The missionary one seems to work kind of okay, but it seems to change the body too much, and there's not much movement.

Also, can you share your training config?

Big fan of your work! After a number of tries, I think lying has better face consistency over kneeling. But still, amazing

Christmas came early 🎄

keep it up, your loras are excellent

the goat just keeps pumping em out

yasssss love you

This one is interesting. On one hand, it works, but also, unlike a lot of the other "Blink" transitions, there's not a huge difference between using a many of the other cum LoRA's and this. The transition can be helpful, but as someone who has made a lot of videos without this.

For example, getting the character into a very different and specific pose is the real problem a Blink LoRA solves. In this case, you can likely already see the character's face clearly, so the amount of transition needed is already within the capabilities of stock Wan.

Use this if the character is in a really dynamic pose.

I also have a minor critique of the cum quality, but I have that issue with every 2.2 LoRA.

So, it's good, but less useful than the other Blinks.

If I can make but a humble request, a blink pov looking back doggystyle would be goated.

This is great, thank you so much for making it.

Excuse the dumb question, but I was wondering what checkpoint or merge models you use?