This LoRA should play well with other LoRAs without needing to mess with strength much.
All videos contain metadata required to generate them.
Here's a sample Prompt (the // are comments, so you can leave those in):
//Simple description of the scene
Two gorgeous women in wedding gowns passionately kissing eachother.
//Describe the characters
They are both thin and gorgeous.
The woman on the right has her hand on neck of the woman on the left. She has her long blonde wavy hair adorned with white flowers. She is wearing a long white lace wedding dress.
The woman on the left has her long red hair in a fancy bun. She is wearing a short silk wedding dress.
//Describe the background/environment
The background a white wedding trellis with flowers. They are in a garden with lush greenery.
//Style of the video
The video has a delightful bright vibe. It is a realistic video.
//Describe the camera angle
long-shot. Realistic.Description
Final version
FAQ
Comments (15)
Thanks! Awesome LoRA. Works well with other characters as well. Good work!
Yay kisses! I've tried several times to improve my AmorousLesbianKisses model but no matter what I try it wants to attach their tongues together. I've tried all kinds of settings, low and high resolutions, really low LR, high LR, high then low, low then high, cosine annealing just... so much. Any thoughts or similar experiences? Regardless, good work!
How big was your dataset, and did you follow a standard for captioning @blyss ? I didn't focus on tongue-kissing, so that is definitely a reason.
I wrote this article to document my methods https://civitai.com/articles/11942/training-a-wan-or-hunyuan-lora-the-right-way
@ComfyTinker Yeah I read your guide it's good stuff and I've definitely benefited, especially from training at lower res! My dataset is currently 35 video clips standardized to 144 frames @ 24fps (I usually just do a head extraction of the first 129 frames to train, 144 is just to have a lil buffer of extra frames in the clips) My captions should be good, I really liked your advice of "caption it like it already works" that's the best explanation I've read for how to do it and is very close to what I was doing anyway, with verbose natural language captions same as I prompt. What I might do (especially since this LoRA exists now) is remove all the non-tongue kissing ones (nominally mine was focused on amorous kissing in general, which might or might not involve tongue kissing) and try to focus specifically on tongue kissing only and see if that helps.
@blyss Feel free to use this LoRA as a base, and just train on top of it if you like.
Could this work with two characther loras?
I'll do some testing in a bit. I think it should, assuming the LoRAs are trained to isolate appropriately (and dont just change all characters in a scene)
Yep! If the character LoRA is trained to be able to isolate the character in the scene (like saying "the woman on the left is <character name>"), then it works. I just tested it with a custom LoRA I'm working on named "Candace Cane" that has blonde/blue ombre hair, and a consistent face.
Once I've fine-tuned that character model better, I'll post some examples
Then it should be possible to get left one one character and right one another and use this lora too?
@zncyibeni419 Yes. I was able to this exactly.
I have been able to try it a bit but noticed it mixes up the persons details a bit, one describing left one with auburn hair and other with blonde but sometimes they both are blonde and such
Sadly, no way to truly isolate the affected areas of multiple LoRAs currently, even if you are explicit with the trigger words and carefully prompt which character you want impacted by each LoRA. It 'kind of' works, at best, but the resulting faces will always have a little blend of both LoRAs. Maybe one day.....
@jm112368767 Character isolation is possible with HunYuan, but requires tricky prompting. The issue is that many LoRAs aren't trained properly and are instead just standard person replacement LoRAs.
@ComfyTinker Would love to see an example and some associated LoRas and workflow...
nice
Details
Files
PassionateKissing_FINAL_HunYuan.safetensors
Mirrors
PassionateKissing_FINAL_HunYuan.safetensors
PassionateKissing_FINAL_HunYuan.safetensors
PassionateKissing_FINAL_HunYuan.safetensors
PassionateKissing_FINAL_HunYuan.safetensors
PassionateKissing_FINAL_HunYuan.safetensors
PassionateKissing_FINAL_HunYuan.safetensors
PassionateKissing_FINAL_HunYuan.safetensors