Anal penetration Backroom casting style
Prompt example:
asss_pen_on_table, The video presents a split-screen view of an intimate act. On the left, a woman with short, reddish-brown hair and visible tattoos on her arms is with her knees on a black chair and she is leaning forward with her arms resting on a dark wooden surface. Her expression is neutral as she looks down. A man's upper torso is visible behind her. On the right, the focus is on the same woman's lower back and buttocks, which are bare. The same man's hand is visible, holding his erect penis and guiding it towards the woman's anus. The man's feet are also visible at the bottom of the frame, on the ground, while the man is standing.
Description
FAQ
Comments (22)
Works well, bad aim =D. But at 1.2GB it is slightly big for a "action" lora. PS: I have had the chance to make one yet, so no clue why it happened.
what's the difference between the two models?
backing time... i guess. 1200 versus 1400.
Looking at your example, the timing between the left and right screens doesn't seem to match; the woman's facial expression doesn't change upon penile penetration?
Try entering this:
Preserve the same characters face and identity in both split-screen view.
Would love to see if it could have a synchronized facial expression upon insertion, the example doesn't show that yet! Can you post one that does?
Love the concept and its worth continuing this as a series. Before you do anymore - I would refine this one so that the front facial response better sync with the action and get the file size down. Good first effort!
I think if you refine the prompt you can manage to match the front face action with the back action. You might need to specify in the prompt that are connected. Try to use an ai tool chat to ask to create for you a prompt that does not break the connection between the 2 side by side screen.
Regarding the file size, it is my understanding that it depends from the rank I use in aitoolkit. The higher the rank, the better details in the lora, the bigger the lora file generated. Imagine i am trainjg / testing as we speak a higher rank = a bigger generated lora size to see if i can refine the details.
This lora has issues but unfortunately I am new at lora training and I am trying to understand what breaks certain things. Like I have issue with the details on yhe male and female genitalia
Need this in i2v man !
It works perfectly fine with i2v
All the t2v I share are compatible with i2v generation.
Thanks, that's work, nice.
did this get deleted?
This is such a great idea. If the split screen motion could be synchronized, and the anatomy improved a bit, it would be a GOAT. On my end the anatomy is still a bit fuzzy. Fuzziness could be because I'm doing i2v with ref2va, though.
Great Lora mate, this works very good with RefMods. :D
could you please share the workflow you used? I would truly appreciate
@nnewtoai719 https://civitai.red/models/2663838/plaguekind-minimax-h3-sparse-attention-refmod-ltx-workflow-eros-or-sulphur-compatible-or-faceid?modelVersionId=3346777
I used this workflow but with minimax_h3_fl2v_turbo_4step_v1.0_768p_comfyui_bf16 at 0.75. 10 steps.
@Smurfypie Thank you so much!
This is absurdly oversized, and you don't need a lora to do this at all.
then do not download it and move on. what is the f point to comment? Share your prompt, work etc so other can copy what you do on this, because people might not even be ale to do the video like you claim you do...
@nnewtoai719 Definitely a needle dick move to say "you don't need a lora" without providing any example of how to do it without a lora.
@NeilHamburger yeah, but I mean:
1) I am learning
2) I am sharing for free and they complain
WTF
Give me a break!
I didn;t foce people to download the lora. If it is shit, whateve, move on if these people do not have any suggestion to make it better. My goal is to becoem better. It is always the same people sharing lora. it like an elite circle. isn;t that fucked up?