To use this Lora, you'll need to find portraits of two people and then use an image stitching tool to combine them into a single picture before using the I2V. When stitching the images, pay attention to balancing the head sizes of the characters and the distance between them—unless you want to see a big head kissing a small head or the video cutting off right before they kiss. You’ve got to set the stage for their kiss.
You can also try using full-body shots since I’ve included training data for them, but I haven’t tested this myself. It’d be great if someone could share examples of it works with full-body images.
The reason I made this Lora is that getting characters to kiss without it—especially when they need to cross the borders of the image while keeping the background relatively stable—is really tough.
I stopped the training process when I felt it was about right and picked the checkpoint with the highest epoch number. For testing, I stitched together some real photos I found online to check the Lora’s performance, but I don’t think I can share those—it might get me into trouble. Still, I can tell you the results were seriously impressive.
trigger word：
“They started looking at each other and moved closer, then they started kissing.”
The example prompts and trigger words were just made up randomly the first time I used them—I didn’t even filter or refine them. You can use your own prompts instead.
The main purpose of trigger words is to help align the pixel vector paths with the motion distribution probabilities during training. For users, they’re not really that important.
Description
FAQ
Comments (16)
Weird, you're saying comfy doesn't have any sort of basic image editing extension?
I use Face Bounding Box (comfyui_faceanalysis), Image Crop/Image Flip (comfyui_essentials), and Concat 2 Images to Grid (comfyui_fearnworksnodes), which basically automates the process of putting the faces next to each other.
This is awesome. If we can use this same method for other 'activities' involving two people, that will be something.
This concept works without the lora. So do other "activities".
@leisure_suit_larry How can do that? like power lora?
@dongwen555886 - I'm saying that you can just use a photo of two different people and prompt for them to kiss... with no lora. It's like the first thing I did with i2v and it works great, no lora necessary.
👍🏻👍🏻👍🏻
This is outstanding, great job coming up with this.
This can perfectly be done with Wan out of the box. It doesn't care about the background if you don't prompt that something's happening in there. Wan sees two heads, you prompt for a kiss, they'll kiss. No need of a LoRA for that.
No more stitching two photos together before you i2v. We've got this for FLUX, if you can get a good one on the terrible awful image generator. https://civitai.com/models/290928/cinematic-split-screen-style-xl-f1d
I think you're missing the purpose of the lora
does it only work with stitched images? not with images that start with two people in them?
Late reply, but it works with two people in the image as well
Can I get this workflow?
Has anyone had trouble identifying gender? Is it more likely that the right side is a boy and the left side is a girl? Is there a solution?
You could edit the file with other apps or programs like premiere, after effects or even moviemaker. Just apply mirror image or sideswap. In each program is different.