I'm just amazed by the face consistency .. ty so much.

Does this LoRA work only with one camera angle?

just fyi, your description has an incorrect title still

How do you avoid the subject's face out of the frame or getting cropped??

Sometimes only half of her body is visible without the face.

You make good models.

Must say this is a very powerful lora

Why doesn't anyone create a model to create an anus for i2v? if there is no anus in the original picture, then the model simply ignores this moment, there is simply no anus in each render. Why doesn't anyone want to solve this problem? Does no one give a damn about this?

Sometimes i got a orgy

I'm not a big fan of this type of Loras, as they are very generic, but this one has great consistency in terms of faces and bodies, and the result is quite realistic. Really impressed.

Thank you for sharing ❤

very good lora!

Thanks, very good lora!

This is great, the Image to Video vs Text to Video makes for so much more fun. Thanks for making it.

Like the Op's other I2V Lora, this Lora makes the face very consistent, the movements are also perfect, even you can add a prompt about the female body type, it can be followed, the only pity is that the female genitals are a little ugly, I have tried many times, many times the female genitals look like a loose hole, I tried to use another T2Vlora combination, it didn't work, I know this is not a first-person Lora, So the model's attention may not be focused on the female genitalia while training, anyway, this one is a great Lora, thanks!  

All your released LORAs are ok, but pathetically inadequate since they are limited to 5 seconds only. Even if you try to generate longer video they simply loop. The LORAs should be made using more varied materials to be usable in longer videos. 5 seconds ANYTHING is woefully inadequate for anything.

How does the dataset look for these kind of loras?

This Lora is really impressive! The transition effects are great, and I'm getting a lot of nice variation in my generations. If you’re open to sharing your training process, it would be fantastic to hear about it.

I have a couple of observations that could potentially improve things:

Insertion Poses: I've noticed that in some insertion poses, the penis seems to already be quite deep inside, and sometimes connects with what looks like a slightly broken or half-formed penis shape. It might benefit from a smoother transition.

Detailing: Increasing the detail/resolution of both the female genitals and the penis itself would really elevate the quality. Perhaps more training data focused on those areas?

Overall, the "Casting" Lora series is fantastic! You're doing great work and I’m excited to see where you take it.

This may be the best lora ive ever used, might have to donate

This is actually fire. Joining, donating, bravo.

Please add second hole :D

I'm loving "Casting Roulette" with a group photo

Please tell me what the dataset looks like, I want to try training a similar LoRa . Please tell me how to correctly create a dataset for similar LoRa.

Any way to put a certain man's face for the guy? That'd be cool if it'd take a pic with a man and woman and then make the video accordingly.

Love your work btw! Cutting edge!

I couldn't help but notice a blowjob casting test lora in one of the embedded workflows from a video. When might this be released? :)

I must say, this is a very interesting LORA. Sometimes it manages to do what cannot be achieved by standard means: skip a certain number of frames and produce the resulting scene described in the prompt. That is, when using the LORA, the model does not try to smoothly connect the prompt with the original image and build a sequential change of plot, but simply skips the unnecessary parts. Have you considered looking at the LORA in this way and training it specifically for skipping, without linking it to doggystyle and the like?

Just a heads-up: this jumpcut-style LoRA might be at some risk on CivitAI, based on recent trends.
Nothing confirmed — but if you’re interested, you may want to grab it sooner rather than later.

Does anyone have any idea how to make loras work sequentially, without having to make two different videos, I want in I2V for it to start with a blowjob, using blowjob lora, THEN screen lights up and goes dark again for doggystyle, for this lora. [SCENE] X,Y,Z doesn't work AT ALL. Would a BREAK command work?

What CFG, what steps? I never know what settings to use with each LORA, and usually get error/static as a result. I tried 1 CFG as someone suggested but it was messed up. Does noise seed matter? I also "start at step" and "end at step" in the nodes and no idea what to put there either. I did use 0.7 strength for both high and low noise, as well as 16 fps, but I never get good results.