Tip appreciated if you enjoy the work, these take a lot of time to create! https://ko-fi.com/misticrain69
Commercial use is now allowed! Just please credit me :3
Update for LTX 2.3. This version works much better than the LTX 2 version
I use the ltx2 furry nsfw lora at a strength of 1 with this and distilled lora at a strength of 0.5
In the included videos (yes it works with anime and non furry the same as it does on furry.) there is the triple sampler workflow I highly recommend using I get much better motion using it vs the stock workflows. Idk if it works on t2v.
LTX 2.3 needs detailed prompting for best results. Be specific about what you want and refer to the prompting guide LTX posted. The prompt I used for the included video is in the metadata just check the video for the prompt.
Update: Uploaded a better lora you should notice more stable motion
A paizuri lora for LTX2. Took a while to dial in training settings but this one turned out pretty good.
Don't use the distill lora at negative strength on ltx 2.3 advice below is outdated and for ltx 2
Workflow is embedded in the attached video. Works great on realistic and animated/anime.
I use it with my motion lora https://civarchive.com/models/2344781/ltx2-better-nsfw-motion and with the nsfw lora https://civarchive.com/models/2310920/ltx-2-i2v-nsfw-furry-multi-purpose-sex-lora all loras at a strength of 1 The liquid LTX-2 Liquid physics lora was also accidentally toggled on for the furry gens but I don't think it had an influence on the output since there was no liquid in the video lol https://civarchive.com/models/2348942/ltx-2-liquid-physics I also use the distilled lora at a -0.4 strength with the distilled model. For whatever the hell reason in reduces any glitchyness I get in gens I have no idea why but I did multiple fixed seed tests with and without the lora. Someone smarter than me may know why but still no clue why. Idk if it works on t2v its made for i2v.
Audio can sound odd at times like with all ltx2 loras.
! If you post things in the gallery that violate the TOS I will block and report you. Don't do this shit.
For 18+ only. This is NOT to be used for any illegal or unethical purposes or on any real person or likeness. Don't be fucking sus.
Description
Booba
FAQ
Comments (6)
Keep going my friend this is good!
Any prompt guidance? What phrase or trigger words did you use in training?
I didn't use trigger words. I really avoided giving it things like "titfuck" so the model didn't have to learn that word from scratch. Also differential guidance when I trained with it enabled really fucked the results. It made all anthro characters gain an uncanny human face lol. When I disabled it I got actually good results. I rented a Rtx pro 6000 runpod and trained without quantization on 768x768 resolution. Heres some prompts from the dataset.
This woman is using her large breasts to stroke a mans large cock, she moves her breasts up and down going the full length of the large cock, she starts at the base of the cock and moves up to the tip, her breasts jiggle as she moves them
A nude furry cat woman with black fur and black hair laying on a light grey floor between the man's legs, the man thrusts his hips as he holds her breasts squeezing them together as his penis slides in and out between them, the perspective is from the man's viewpoint
A dimly lit cave with grey rock and a pool with green water, a fully nude furry bunny woman with light brown fur and white hair moves her large breasts up and down the full length of a furry man's penis, she closes her eyes, the side-angle view captures the action
With her large breasts the anime woman stokes a mans penis, she quickly moves her breasts up and down going from the base of the penis up to the tip
POV Blowjob and cumshot pleasssss
Maybe if no one else does it. LTX2 is pretty expensive to train. Had to rent an RTX pro 6000 to train this one.
I noticed your example (and workflow) uses Kijai's LTX. Does this still work with distilled Lora? Because I see several workflows stating it needs to be the dev version (which if using the video comes out scrambled, or cites dimensions error)