You can make money with my models. Sell what you generate, use it in paid work, run it on paid services, merge it, sell the merge all of it, no permission needed. Credit is appreciated but never required. New releases might have a short early access window behind Buzz, then go free and free is permanent. Once a model is free, I will never re-paywall it or pull it for a paid platform, and you're welcome to mirror, archive, and rehost it so it stays free no matter what happens to any one site.
Feel free to mirror any of my loras to HuggingFace
Use this model for anything you want, commercial use included, no permission needed, just credit me. The only restrictions are the common-sense ones in the TOS at the bottom of the page (NO illegal content, NO minors, NO real people). Just don't be sus.
This lora was trained exclusively on 2D animations. It was trained on most of the common NSFW actions like fellatio, cowgirl, missionary etc. It was also trained on a lot of cum so it is good at cream pies and oral cream pies out of the box as well as just cum in general. Be specific about cum or else it can appear on parts of the body you did not ask it too! The lora can be used on 3d animations as well but they will look choppy and low FPS since its trained on 2d animations that were animated with on 2's and on 3's. Trained on Sulphur 2 base and I use 10eros to generate the videos with the lora since 10eros is the i2v optimized version of Sulphur 2. I also always use my VBVR reasoning lora with it to generate videos since this gives the best prompt following in my experience. You need to prompt for the emotions you want you characters to have or they will typically have the same expression as they did in the input image. Videos were generated in 540p and 720p and upscaled I will try to upload 1080p generated videos later. This should also be able to do anime/non furry videos as well since it was trained on part furry and part anime but I have not tested it yet since my focus is furry.
Prompting is pretty typical and for some reason you can prompt in the first person sometimes and it can give a nice result for things like fellatio. Dataset was mostly equal parts furry and anime/animated NSFW videos. Example videos mainly show creampies/cum in mouth but yes it works perfectly fine without cum! This lora took entirely too long, around 70 hours of training dialing in settings I am glad it finished properly. This version is sadly not the GRPO RL version since I could not get the RL version to stop getting reward hacked and messing up the color. So until musubi fixes the RL or unless I can figure out how to get past the reward hacking there wont be a RL version, sadly. For anyone who cares about training details the lora is rank 128. I tried rank 64 but it had no where near enough capacity to actually encode the 2d look and cadence for some reason. It was trained on 370 videos. For future versions I plan to train on close to 1k or more but not sure when that will happen considering how hard this version was to make and how much time it took.
Disclaimer & Terms of Use
This model is provided "AS IS", without warranty of any kind, express or implied, including but not limited to the warranties of merchantability, fitness for a particular purpose, and noninfringement.
IN NO EVENT SHALL MisticRain69 BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH USE OF THIS MODEL OR ANY OUTPUT GENERATED WITH IT.
18+ only. You, the user, are solely responsible for anything you generate with this model and for ensuring your use complies with all applicable laws and the CivitAI Terms of Service.
You may not:
produce anything illegal, non-consensual, defamatory, or used for harassment, deception, or fraud;
create any content depicting, or appearing to depict, minors (real or fictional) in a sexual context — no exceptions, ever;
use this model to depict any real, identifiable person — celebrity, public figure, or private individual — in any context, SFW or NSFW. No likeness use, period.
If you post outputs to the gallery that violate the CivitAI TOS, you'll be blocked and reported. Don't do this shit. Don't be fucking sus.
Description
FAQ
Comments (17)
Gosh and this was needed. Hopefully it exceeds the other furry lora which had issues with ghosting. I'll give this one a try as soon as I can after work.
LTX 2.3 in general has issues with some ghosting when the movement is fast.
Also are you able to see the posted videos currently? I dont see the XXX rating on my side so im worried civit is borked like usual and no one else can see the videos but me.
@MisticRain69 It appears to be borked. No videos on my end.
As for ghosting. You're right that LTX can have a bit of a problem with fast movement, but thats mainly smearing, which is often solved by generating at a higher FPS. But this was a different issue, during movement you would see a sort of after image of where the character had previously been. And it only happened with that Lora, it was very obvious. (Multi-purpose furry sex Lora) Its a good lora and it brought furries and 2D to life in a way that LTX normally couldn't. Even more so for Sulphur which is mainly trained on humans and can't do 2D very well. The ghosting was very unfortunate and did spoil the overall quality of the clips.
@OhNoMyChickens Interesting i never had that issue on 10eros using that lora in wan2gp.
@OhNoMyChickens Hopefully civits rating bot will stop breaking so people can actually see the vids. Civit mod said its pretty broken rn.
@MisticRain69 Could also be a workflow issue. Would love to solve the problems since it's a good lora that can be used for a wide range of things. But weirdly enough it's like one of maybe 2 or 3 Loras that do this while the others work just fine. As for Civit being broken, it's almost the weekend so it's par for the course at this point...
@OhNoMyChickens Civit just unborked so vids should be up
Been testing it a lot and other than the smudging you get at low FPS with LTX. It works very well. Good voices, good movement, hair and tails are brought back to life. I suspect there is some MLP in there because Twilights and Applejacks voices are spot on.
You said you used 370 videos. Is that 370 individual 5 second clips or 370 full videos, which were then sliced into clips? How were they captioned? How do you know what to look for? Im interested in making a similar Lora someday. 2D on LTX is in a rough spot so this Lora helped a lot.
i hope this is the silver bullet for 2D conditioning for ltx eros/sulphur
I made a video post with some example the other day. It works very well overall. It adds physics back to tails, hair and other parts for 2D images. I do believe it's a bit of a shame the choppy framerate (which is part of the style) isn't somehow adjustable, since LTX does have some issues with smearing at lower framerates.
@OhNoMyChickens The framerate is interesting because all videos used are 24 fps and not super choppy its just how LTX learns 2D I guess.
@MisticRain69 interesting. So maybe it's just a prompting issue then. Even stock LTX can sometimes give you a choppy framerate on 2D things. As much as the choppy framerate is nice for emulating that animated look, the smearing can be distracting sometimes.
Please provide the workflow you used to generate these demo videos. I can't get videos of similar quality.
I'm pretty sure he uses the basic RuneXX workflow which is found here: https://huggingface.co/RuneXX/LTX-2.3-Workflows/tree/main
Specifically LTX-2.3_-_I2V_T2V_Basic.json
It's about as simple as it gets for LTX 2.3, I use it all the time but have modified it to some degree.
I use wan2gp so I dont have a specific workflow. I use 10eros bf16 with the normal 1.1 distilled lora at 0.6 strength and my reasoning lora at 1.0 strength with the 2d lora at 1.0 strength.
Did some tests and the choppyness makes the videos almost unwatchable. Not great