I am admittedly kind of a noob at this, but if you enjoy my humble work, any donations are greatly appreciated: https://ko-fi.com/insult_to_ninjary

Your generosity helps me stay motivated to keep improving and to keep trying new things!

Description:

This LoRA is intended to produce that distinctive jerky and jiggly animation that plays when you tap on your favorite waifu in your favorite gacha game. It got the jerkiness of the animation down pretty well, but I was hoping that the jiggle would've come out a bit stronger. Regardless, it's still pretty decent, maybe even good if you prompt well and get good seed luck. It may also be possible to get more jiggle with better prompting or upping the LoRA strength; I haven't tried a ton of variations yet.

Please note that I will not be making Wan versions of any of my LoRAs at the the present time. I can barely run Hunyuan Video via FramePack-Studio on my low-VRAM laptop. Running Wan in any sort of practical way is probably not in the cards for me without getting a new computer. However, if anyone wants to try to make their own Wan version of any of my LoRAs, I am more than happy to share my training datasets. Just ask!

How to Use:

Full disclosure: my experience with the LoRA is entirely based in FramePack-Studio. Your experience may vary depending on which platform you're using.

The trigger word is g4ch4w41fu, but I'm not sure if it actually does anything or not. Probably doesn't hurt to just throw in there though. More important are these two sentences: She is doing a Live2D character animation. Her breasts are jiggling in reaction to her movement. These were used to capture the main movements, though as mentioned before, the jiggle part can be pretty hit or miss unfortunately.

My general rule of thumb with all Hunyuan Video LoRAs in FramePack, this one included, is to start at a strength of 0.8 by default, and then move up or down as necessary. Using this LoRA at a strength of 1 (or even 0.9) does seem to improve the motion and jiggle a bit, but it can also introduce other problems. But not always! So play around with different strength values!

Below is a sample prompt that I used for the first showcase video. The second showcase video uses almost exactly the same prompt, just with a different character description. I don't know how rigidly you need to follow this prompting structure for the LoRA to work well, so feel free to experiment by changing, adding, and/or removing sentences! Let me know how it goes!

g4ch4w41fu. A cute anime girl is standing in the center of the frame. She is doing a Live2D character animation. The girl has shoulder-length dark brown hair in twintails and deep blue eyes. She has medium breasts. Her breasts are jiggling in reaction to her movement. She is wearing a sexy business outfit. The background is an office. cowboy shot. anime style. The girl is doing an exaggerated victory celebration action. Arm movement is very slow. Hand movement is very slow. She has a slim body. She is smiling, and she is very slightly blushing. Movement is slow, smooth, and steady. The girl's facial appearance generally does not change. The art style does not change. Image texture and color remain consistent always.

Please note that both showcase videos are i2v. The same general prompting structure should also work ok for t2v, you just might need to be a little more descriptive with some things. The quality seems much better with i2v though, so I recommend going that route if you can. If you're wondering why my showcase videos might look a little crisper than yours, it's because I always run a 2x upscale with RealESRGAN_x2plus and 2x frame interpolation with RIFE after generating. I strongly recommend it!

Some Notes:

• The anime style is pretty burned into the LoRA. I think it might be hard to prompt for realism, but let me know if you find a way! You can, however, use a realistic image for i2v, and it should work pretty decently for 2-6 second videos before anime qualities start creeping in too much.

• As is the nature of a lot of these fan-servicey gacha games, the character outfits in the training clips leaned pretty skimpy. Thus, for t2v, unless you prompt clothing specifically, you're going to get a lot of stuff that looks like a swimsuit or other kind of skimpy clothes.

• I included various breast sizes in the training clips. I thought I did a pretty good job of keeping it balanced, but the LoRA seems inclined to generate big boobs, sometimes even if you prompt for something smaller. Just try again on a different seed, or adjust your prompt a bit. It is possible to get various sizes, but sometimes you have to fight for it a bit. Or just use i2v.

• I had a few instances where t2v generated completely bare breasts unprompted, which is weird, because there was no nudity whatsoever in the training clips. I noticed this happened more often when I didn't specifically mention a shirt in my prompt. So maybe describe what kind of shirt you want, if you want to make sure your t2v is SFW.

• A lot of these jerky gacha animations have really fast-moving arms and hands, and this carried over into the LoRA. It can be hard for Hunyuan to keep up with movement that fast, especially if you're using MagCache or TeaCache for a speed boost. So I usually throw Arm movement is very slow. Hand movement is very slow. into my prompt to help out a bit.

Where are v1 and v2?:

v1 was made almost a year ago, when I knew what I was doing even less than now. It kind of worked, but it was very difficult to prompt, the quality wasn't great, and it was generally just not that pleasant to use. I never intended to share it on CivitAI.

v2 was made just a few days before v3. It works pretty well, but I decided I wanted to tweak a few things and retrain. v3 seems like it does everything v2 does, just a bit better, so there's kind of no point to share v2 now. Still, if for some reason you do want to try v1 or v2, I have made them available in my Huggingface backup.

Outro:

If you make anything awesome, please consider posting them to the gallery! That goes for all of my models! I really like seeing what kind of stuff you all are able to make!

If you have any questions, feel free to ask! I will try my best to help! Also, if you have any ideas for future models, please don't hesitate to share them with me!