DR34ML4Y by DR34MSC4PE
This LoRA, incredibly, manages to handle ALL of our specialty LoRA concepts - Double/Single Blowjobs, Missionary, Reverse (and regular) Cowgirl, Doggy, wrapped in one model! It obeys very well and is largely extremely accurate to the given keyword/position! Don't be shy; give it a try! =)
===========================================================================
Update - Due to civit making sure we do not get paid for the use of our model via generation any longer, on-site generation has been switched to 100 buzz per generation. Please voice any concern directly to Civit AI's support over their change effective 11/1.
10/15 - H3 PREVIEW: we're so back.
Note: this is an early preview - expect some T2V challenges. H3 is not WAN/LTX. Prompt accordingly.
Turbo LoRa moves latent space away from NSFW concepts. Results, especially audio, will get worse the more turbo you combine into the mix.
AS ALWAYS: Enjoy, generate responsibly and share what prompting tricks you learn along the way!
==========================================================================
Note - Missionary and cowgirl can have some mix-up if you don't specify the woman is on her back in the prompt.
===========================================================================
DR34MSC4PE is
c0ur4ge (trainer/qa/inference code) /
ERA5ER (trainer/captions/tooling code/data/qa)
Please generate responsibly!
Description
Fine tuning and balancing based of v1 - originally this was slated to release as 1.1 but it doesn't do the improvements justice so here's 2.0. Training continues so stay tuned! NOTE: The samples with bad audio were me testing abliterated encoder - we do not recommend it's use, YMMV!
FAQ
Comments (73)
Best to ever do it. Incredible as always, I can't imagine the effort and dedication that goes into making this... an insane amount of appreciation for your work, thank you!
You're super welcome! We do it for y’all - we do it to learn!
Thank you for the update! Am I missing the paragraph about the revision changes? What is new? Thank you!
TLDR; It sucks less than v1.
Its dataset is balanced a bit more and it's cooked a bit longer. Training continues still as we add data in and, for future versions, add more ultra-wide resolution support. Right now, 1280x720 and up can cause some artifacts in male anatomy and physics, etc.
@c0ur4ge fucking goated
are the trigger words required, or just optional?
YMMV - they're likely to be helpful if you're doing very little mechanical explanation of what's happening, particularly in I2V. In T2V, I'd still recommend them.
Where are the workflows? your video examples did not embed them (why?), and https://huggingface.co/RuneXX/LTX-2.3-Workflows/tree/main has a bunch of unrelated/generic workflows - I don't see anything specific to this LoRA
This is going to come as a surprise to you but I made all my examples with these "unrelated/generic" workflows by adding a lora loader after the model. You know, like every workflow ever in comfy since the beginning of time.
Take some vitamins and try harder. Or download someone else's workflows from literally any image below if you can't be bothered.
If you're hoping to just put "guy fucks woman" into a box and click go, this model isn't for you, my friend.
@c0ur4ge Fucking G.O.A.T.
@c0ur4ge remind me so much of old stack overflow days 🫠
Hey, I think he means that someone could give me a link to download this workflow, which would help me get started :) but since he doesn’t share anything in his folder himself, I’m not going to be the one to help him out—I don’t like selfish people who help themselves and then criticize others
I've gone through a 9hr course start to finish on ComfyUI along with dozens of vids on how to customize workflows. LTX's workflow is like going through a hoarder's cabinets full of tangled christmas lights, and "just loading a lora loader after the model with .6 strength" isn't enough. It would be nice to just have the teeniest tiniest guidance instead of being an uptight gatekeeping coomer on day 3,000.
Could you please create it in a form that is friendly to the Eros model?
Not really? We don’t intend on training against any base other than official checkpoints.
You can try it at lower strengths but YMMV.
I have a low vram of 8BG and that's why I use sage attention. does it affect my output. I need some help with this. any document guide for settings and prompting
sage shouldnt affect badly on gens
How can I get the cum from dripping from the guy's mouth when he cums? I've tried "only cums from the tip of the penis" and it still comes out of his mouth. Any suggestions?
Wait, isnt that normal?
@redlittlerabbit let me rephrase, or perhaps you were being facetious, but "how can I stop the cum from dripping from the guy's mouth as it also drips from the tip of his penis?"
Set NAG to 1+ and play with negative prompt.
@IAmNotABot4Real Oh you fell straight into my trap
@Dariuzs NAG?
@IAmNotABot4Real In Wan2GP on general section, there is NAG setting, set it higher than 1.0 to enable negative prompt. If you use comfy then i dont know.
@Dariuzs Ah, okay. Thank you. I do use ComfyUI.
Testing with LTX 2.3 dev + distill lora. v1 works great (thanks for that!), but v2 create some body parts horror. Just changing to v1 lora with equal params and seed give good result again. Any advices?
The motion is significantly better but the audio becomes bad in I2V.
Which text encoder and checkpoint combos?
any recommended high/low lora strength to achieve fast movement? thanks in advance.
Thanks
the "download" button in the i2v low and high says file not found
It seems like all of them are doing that?
Well this is some bullshit - I've opened a ticket with Civit. Hopefully the fix their shit and quickly.
Seems fixed for now ....
this is common. just refresh the page and try again.
hi everyone! does anyone try for creampie or cumshot on this model? how can i do that?
It's mainly a matter of luck but you can try to have a good starting image first and instead of promoting "she" Ejects out cums from her p.. You can try "her p.. " Ejects out white cums.
@Bellglua thank you man.
Edit: Mate when use your prompt it was like she is blow up cum from her p.. lol. I am trying 'leak out' instead of eject out. I will notify the result.
Edit 2: Mate the best choice is 'he is cumming inside her p.. with his d..'. This one is the best.
@kkorcan72407 I think use drips like her p... drips out white ..
My video is blurry,on wich strengt should I lora use?
curious too
@bullshiz Which model and lora version are you trying? Which workflow? etc etc etc
@c0ur4ge I fixed it with another distilled lora model
@Peacer0342 Would you mind sharing what you used? I have the same problem.
will this be supported in Ltx 2.3 version
Is anyone getting good results in Wan2GP (t2v)? It's just a complete mess for me.
i second that
hmm trying but man the vagina is messed up big time, no matter what value I set lora loader to. The Wan version.
model not found error
This is a transient Civit thing that comes and goes it seems...
There seems to be a problem for male voice generation. Lora blocks it, maybe possible. Check for this, please.
Possible there isn't much in the dataset overall in the shorter snippets used. I'll check into it - that said I've heard male voices before.
THX....works fine for me
Can somebody enlight this newie user, if I use this lora, should I also use the HIGH and LOW versions for I2V present also here? or this one covers it all?
Hello, i would like to know any kind of tutorial to work with this kind of generations, like from the begining, I can't find any advice. Thank you in advance :)
Hi, I'm new to video gen, and tried to get this Lora working, but it's extremely blurry. I tried replacing the Lora on a default ComfyUI I2V Template. The WAN 2.2 one (The white dragon warrior...).
I tried to load a workflow via the videos below but they just getting added as a load video node.
So could someone please share their workflow working with this Lora? I'm overwhelmed by all the different models/Loras etc. and probably just overlooked something simple. I want to learn, but it's easiest to have a working starting point.
For q4km version of ltx distilled, I used distilled lora 384-streght 1 for first 8 steps and distilled lora fro90_ceil72 strenght 0.4 for the last 3 steps.
LTXVImgToVideoInplace
TypeError: cannot unpack non-iterable NoneType object
This is on one of the Rune workflows provided in the link. Anybody know anything about this issue? Crashes after reaching LTXVImgToVideoInplace node...
Hey guys, I have a question. How did you manage to make so many great NSFW videos in the portfolio showcase? Is it because the i2v images are the first frame of the video? Why are my videos so rough? The people are completely inconsistent.
A lot of times you gotta stack other loras with it with different weights to make it look nice, like some other positional thing, or style or whatever. Its more of an art than a science.
@dirkzen We never do this in the portfolio - this is just our LoRA and straight T2V. Just good prompts!
@dyi486788 your videos are rough because ltx 2.3 is trash lol
@mjh02111964582 It's not that bad, it's different (and definitely more challenging to train) but if you're a person who's only ever used the distilled workflows, I'd recommend just using dev.
16-bit formats (FP16, BF16) and Q8_0 GGUF have better quality than plain FP8. If you're using an even lower quant then quality will likely take a bigger hit. If you're talking about Wan 2.2 then the different speed-up loras can have a big impact on the quality. The very old 256-rank distilled ones tend to good unless you need more motion in which case there is a new 256-rank lora that was released. Everyone wants a free lunch (speed-up and lower VRAM by using heavy lossy compression quants plus speed loras) but there isn't one.
Only reason to use ltx is because natively you can make videos longer than 5 seconds.. i wish wan could do that without having to make multiple videos using last frame.
@laboon201 Have you tried SVI 2.0 pro for Wan 2.2? I tried to use it but the early workflows were inscrutable. I wonder if those have been made more usable by now.
Does anyone have a good WAN2.2 i2v prompt that starts with a man and his erect penis, and a woman enters the scene and sits on it? Focusing on the man's facial expression. I'll send some buzz!
Perhaps this could help, can be used in i2v...
https://civitai.red/images/127381238
@TieFighterPilot Thanks! I'm going to try it!
@IAmNotABot4Real Thabks for the buzz! 🫡
Details
Files
DR34ML4Y_LTXXX_V2.safetensors
Mirrors
1811313_dreamlay_ltx_V2.safetensors
DR34ML4Y_LTXXX_V2.safetensors
DR34ML4Y_LTXXX_V2.safetensors
DR34ML4Y_LTXXX_V2.safetensors
DR34ML4Y_LTXXX_V2.safetensors
drmly2.safetensors
DR34ML4Y_LTXXX_V2.safetensors
1811313_dreamlay_ltx_V2.safetensors
DR34ML4Y_LTXXX_V2.safetensors
DR34ML4Y_LTXXX_V2.safetensors
DR34ML4Y - All-In-One NSFW v2.0 - LTX2.3 - m15510n4ry,bl0wj0b,d0ubl3_bj,d0gg1e,c0wg1rl,r3v3rs3_c0wg1rl.safetensors
DR34ML4Y_LTXXX_V2.safetensors
DR34ML4Y_LTXXX_V2.safetensors
DR34ML4Y_LTXXX_V2.safetensors
DR34ML4Y_LTXXX_V2.safetensors
DR34ML4Y_LTXXX_V2.safetensors
DR34ML4Y_LTXXX_V2.safetensors
DR34ML4Y_LTXXX_V2.safetensors
DR34ML4Y_LTXXX_V2.safetensors
DR34ML4Y_LTXXX_V2.safetensors
dream2.safetensors
LTX23_DR34ML4Y_v2.safetensors
DR34ML4Y_LTXXX_V2.safetensors
DR34ML4Y_LTXXX_V2.safetensors
DR34ML4Y_LTXXX_V2.safetensors
DR34ML4Y_LTXXX_V2.safetensors
DR34ML4Y_LTXXX_V2.safetensors
DR34ML4Y_LTXXX_V2.safetensors
dr34ml4y-all-in-one-nsfw-wan-ltx2.safetensors
DR34ML4Y - All-In-One NSFW v2.0 - LTX2.3 - m15510n4ry,bl0wj0b,d0ubl3_bj,d0gg1e,c0wg1rl,r3v3rs3_c0wg1rl.safetensors
DR34ML4Y - All-In-One NSFW v2.0 - LTX2.3 - m15510n4ry,bl0wj0b,d0ubl3_bj,d0gg1e,c0wg1rl,r3v3rs3_c0wg1rl.safetensors
DR34ML4Y_LTXXX_V2.safetensors
DR34ML4Y_LTXXX_V2.safetensors
9a4bf387-b754-4820-9901-b56d517a5492-DR34ML4Y_LTXXX_V2.safetensors
aa663714-30d0-41b7-a2ec-fddabe51be4b-DR34ML4Y_LTXXX_V2.safetensors
e4e63c70-fb76-4dd0-bb2b-5ff9813f7c6f-DR34ML4Y_LTXXX_V2.safetensors
DR34ML4Y_LTXXX_V2.safetensors
DR34ML4Y_LTXXX_V2.safetensors
DR34ML4Y_LTXXX_V2.safetensors
DR34ML4Y_LTXXX_V2.safetensors
DR34ML4Y_LTXXX_V2.safetensors
DR34ML4Y_V2.safetensors
DR34ML4Y_LTXXX_V2.safetensors
DR34ML4Y_LTXXX_V2.safetensors
DR34ML4Y_LTXXX_V2.safetensors
DR34ML4Y - All-In-One NSFW v2.0 - LTX2.3 - m15510n4ry,bl0wj0b,d0ubl3_bj,d0gg1e,c0wg1rl,r3v3rs3_c0wg1rl.safetensors
DR34ML4Y - All-In-One NSFW v2.0 - LTX2.3 - m15510n4ry,bl0wj0b,d0ubl3_bj,d0gg1e,c0wg1rl,r3v3rs3_c0wg1rl.safetensors
DR34ML4Y_LTXXX_V2.safetensors
DR34ML4Y_LTXXX_V2.safetensors
DR34ML4Y_LTXXX_V2.safetensors
DR34ML4Y_LTXXX_V2.safetensors
DR34ML4Y_LTXXX_V2.safetensors
DR34ML4Y_LTXXX_V2.safetensors
DR34ML4Y_LTXXX_V2.safetensors