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
We're trying something new for Wan 2.2 - so here's an early gift from we to you!
FAQ
Comments (42)
Oh boy, the goat's back in town. Thanks for this, mate, will take it for a spin
Update: yeah, as expected, this thing is a BEAST. To me, preserving likeness is always top priority, and this thing delivers. Huge thanks for your work - will post a few samples.
Cheers - we hope you enjoy! ;)
This Lora is amazing, i use for i2v blowjob, without any nsfw fix lora help it can do deepthroat action perfectly!
the motion is amazing! great work!.. im struggling to keep the original faces clean on i2v... any loras to combat this anyone?
Try mixing in one of our other WAN Loras!
ERA5ER yeah ill have a play around, thanks :)
sirbazzrickk We MAY try an I2V version of the tech preview but it depends on our schedule and if we can spare a few days to train it before Wan 2.2 is out. We’ll see if we can get a 2.1 tech preview out but ultimately we want you to see “the real deal” soon.
c0ur4ge Wan 2.2 will be tomorrow. But take your time
Very nice!
The best! Absolutely goated motion loras. Looking forward to what the future will bring :)
Any advice on running this with lightx or FusionX? especially the missionary and cowgirl scenes have almost 0 movent with the lora
I’ve honestly never tried it as it’s not what we’re training on as a base - that said, I know that I2V variants we’ve done have been reported to work fine with them so it may be a matter of parameters.
I’ll let others chime in if they’ve had success!
All of my generations have been using lightx, mostly at 4 steps. I also use the MPS reward lora.
sharp3 Thank you, i will study your workflow now, thanks for leaving it in the generation. Do you work with 2 workflows, one that makes an image and one that uses i2v on that?
craftbin583 Yes. I make my images in ForgeUI, mostly with various pony checkpoints.
This lora improved my SFW human with human prompts by 1000%
This is perhaps the most surprising comment so far lol! Please share! :)
i'm surprised how good this lora works.
thanks for sharing
we were surprised too! More to come!
Marvelous lora.
Wish it had more diverse dataset and without these TRY_T0_R3M3M63R_TH3_T4G.
That’s, I’m afraid, exactly what makes it work. :)
A new version is cooking with some additional positions for WAN 2.2!
FANTASTIC JOB! Congratulations and thank you very much! This is one of the best NSFW loras and the best blowjob for WAN 2.1. Please make a version of this for WAN 2.2??!! That would be amazing and revolutionary. Thank you very much in advance.
it works for 2.2 already my friend
fadog74502638 And there’s a version for WAN 2.2 5B in the oven as we speak.
fadog74502638 Yes, my friend. I can get it to work on WAN 2.2 I2V, but my main interest is WAN 2.2 T2V. I can also get it to work on WAN 2.2 T2V, but it doesn't compare to the quality of LoRas trained exclusively for WAN 2.2 T2V. When using a T2V 2.1 LoRa on T2V 2.2, it loses all the incredible features of the "high noise" model; only the "low noise" model is practically used. I tested a single LoRa trained exclusively for WAN 2.2 T2V, and the quality is FANTASTIC! It looks like real life; the lighting, movement, resolution, and skin textures are impressive, something that doesn't happen when using LoRas from version 2.1 T2V. I think developers should now dedicate themselves to creating LoRas exclusively for 2.2 (I2V/T2V) to reach another level of excellence.
rafaaguiar990170 Yeah - those will come after! 5B first just because I’m very surprised with the quality and it’s the fastest to train. 14b(s) are next !
c0ur4ge Great! Thank you so much for your effort and great work for the Wan community!
IMPORTANT: INFORM ALL WAN USERS! I DISCOVERED HOW TO USE LORAS 'NSFW/SFW' IN WAN 2.2 T2V, GETTING THE FULL POTENTIAL OF "HIGH NOISE" AND "LOW NOISE." THE RESULTS WERE IMPRESSIVE!! I ran it on an RTX 3060 6GB, 32MB RAM, 480x480, length: 41, generation time: 5 minutes to generate each video. The quality is fantastic! For those of you with more powerful computers, the results will be even more incredible. I used the models: WAN 2.2 - Q4_1.gguf FOLLOW THE STEPS BELOW: 1: Don't use acceleration/enhancement LoRas (such as Lightx2v or other) on "high noise"; use only on "low noise" with strength 1. (I'm using this version of Lightx2v LoRa): lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank64_bf16 2: Use 'NSFW/SFW' LoRas on both "high noise" and "low noise" models. On "high noise," use strength 2, and on "low noise," use half the strength, with strength 1. You can vary the strengths as you like, but always keep the "low noise" strength at half the strength of the "high noise." 3: Use these settings in KSamplers: "high noise": steps: 9 / cfg: 3.5 / euler/simple / start_at_step: 0 / end_at_step: 1 "low noise": steps: 9 / cfg: 1.0 / euler/simple / start_at_step: 1 / end_at_step: 10000 Note: To improve speed, I set the Windows system to "best performance," disable unnecessary background programs and applications, and lower the desktop resolution. This significantly speeds up video generation. I hope this helps.
What does your lora training settings and dataset look like?
We don’t stray far from recommended settings on diffusion pipe - as for the dataset on this one, it’s evolving. As you can gather its most of our previous concepts in one configuration with some balancing.
ugh the loras are soo good. hopefully you can add some gay capabilities too (ie anal t2v or make a separate lora)
It’s interesting you mention this because we have a dataset that is mostly just male anatomy but I wonder if it would help flexibility there - we’ll consider it!
c0ur4ge cool! would be cool to see. thank you!
this at strength 0.85 and using the Self-Forcing / CausVid / Accvid Lora, massive speed up for Wan2.1 made by Kijai lora on strength 0.7 is amazing! keeps the faces true, helps the lighting and movement... excellent work
Wild. Need to try this myself even. :)
If it helps anyone on 2.2 T2V, I think the core of 2.2 is in the High Noise model, especially the movement. The low noise model is pretty much 2.1. You want the high noise model to do the bulk of the video, and have the low noise one just do clean up duty. If you let the low noise model do too much work, it outputs a 2.1 style video, its hard to explain. And as someone else said, the high noise model is at its best without the acceleration LORAs, its needs CFG to be higher than 1 to shine.
cant wait for the final version
I’ve got it trained but not I have to evaluate and interpolate of the best of the 5B options for v1. Shooting to have it out by this weekend sometime!
I seem to get error ' not an mp4 file' at the end of generating and it scraps it.... is it loras not playing nicely with eachother or down to settings/files.... its infuriating and any help would be great!
This is definitely something in your workflow, m8.
Details
Files
DR34ML4Y_AIO_PREVIEW_V1.safetensors
Mirrors
DR34ML4Y_AIO_PREVIEW_V1.safetensors
4_t2v_nsfw-all-in-one_V1.safetensors
DR34ML4Y_AIO_PREVIEW_V1.safetensors
DR34ML4Y_AIO_PREVIEW_V1.safetensors
DR34ML4Y_AIO_PREVIEW_V1.safetensors
DR34ML4Y_AIO_PREVIEW_V1.safetensors
DR34ML4Y_AIO_PREVIEW_V1.safetensors
DR34ML4Y_AIO_PREVIEW_V1.safetensors
wan21_14b_t2v_nsfw_lora.safetensors
wan2.1-t2v-s-all-in-one.safetensors
DR34ML4Y_AIO_PREVIEW_V1.safetensors
dreamlay-nsfw.safetensors
DR34ML4Y_AIO_PREVIEW_V1.safetensors
DR34ML4Y_AIO_PREVIEW_V1.safetensors
DR34ML4Y_AIO_PREVIEW_V1.safetensors
DR34ML4Y_AIO_PREVIEW_V1.safetensors
DR34ML4Y_AIO_PREVIEW_V1.safetensors
DR34ML4Y_AIO_PREVIEW_V1.safetensors
DR34ML4Y_AIO_PREVIEW_V1.safetensors
wan21_14b_t2v_nsfw_lora.safetensors
DR34ML4Y_AIO_PREVIEW_V1.safetensors
DR34ML4Y_AIO_PREVIEW_V1.safetensors
wan21_14b_t2v_nsfw_lora.safetensors