Unless he publish the list of loras he used, you are encouraging him to continue stealing other's peoples work...

@K3NK Not sure what you mean. I thought all this AI stuff is built off of skimmed data anyway. Are you saying he generated training data by using someone else’s LoRAs to generate clips of what he needed? That would mean he put his own work into it. But do you know if that is what he did? Even if he trained it all on video clips that were downloaded from this site or from a non-AI site it would be stealing of a sort, but his own work too. So who knows?

@MGHerder i understand, IA is just mixing others ppl art or work, but since Civitai has rules, is basic courtesy to attribute the original authors you're using, that's on a minimum, training on specific sex loras and not even linking the sources, that's shady as f... He is not training on clips, he is just merging the loras without permission nor credit..

@K3NK Oh, I see now. Still a n00b here, senpai. Didn't even realize you could do that.

This is a lora, dude.
Not a diffusion model.

@LFAA ah fuck, im an idiot.

@DigitalPastel you caption different style videos and so?.. tell us how you captioned 200 diferent themed videos, we all will love to do that.. please tell us what too you used.

- "I made over 200 videos using img2vid" (and loras, right?) i tried myself to do an AIO lora from scratch in a single training and is really hard, you telling me you found the way? just with this simpe keyword trigger? XD oral sex, vaginal sex, anal sex, handjob, sex from behind, deepthroat, titfuck, cowgirl position, reverse cowgirl position

i will download it now and i will try generating a penis out of the blue, if your lora can do that, then you prolly used mine loras + all the others.. you must be the best trainer in the world to be able to train an AIO at first try with just videos ...

you also told me there were no passionate kissing lora in the merged model when the final video looked THE SAME XD
anyway.. the issue here is not that much that youre using other loras.. is that you dont give a fuck about crediting the loras you used..
my datasets are mostly real videos (i wont be crediting those, they are outside civitai's ecosystem...), i never trained with IA images, only the face loras i made coz i can control how the character is looking..

@DigitalPastel  what would i know right... i just have trained more than 38 loras in huny, wan2.1 and wan2.2 all local, now tensor.at makes the best loras arround even better than diffusion-pipe

plain and simple: what we want is for you to credit the authors of the loras you used for this.. also on your merged models.. even if you trained on videos as you said.. credit the authors of the loras of those videos.. dont you think? thats much more easy than traning an aio lora with 20 different nsfw actions.. you charge only 200buzz for such a hard work? XD

@DigitalPastel "Do you want me the put on the description what loras I used to make the videos that than were used on the training? Again - thats not a problem AT ALL! Dude, you just needed to ask. omg."

The thing you seem to miss: Nobody should have had to ask in the first place. It's simply the descent thing to do and cost nothing.

The problem is you seem to double down and feel attacked rather than go "Oh shit, you are right, forgot about that", produce links in the description and move on.

This is not a hill you should be willing to die on.
Cooperation goes way further than anything else. Think "open source mentality"...

I was open to believing that you trained this on AI-generated videos until you said you only trained for 5 epochs. Anybody who has trained Wan LoRAs knows that 5 epochs is not anywhere near enough to get passable, let alone good, results. Especially on a dataset containing 200+ videos showcasing many different concepts. It seems more likely that this is a simple merge.

@LocalOptima exactly my point.. i hope his 16k follwers start following us too at least.. xD

Meh idk , doesn't look like just a merge , personally I don't care.
You can use it on gray image and apply more str to see what it does with same tags you had on your lora and will see result if it looks different (no scientific method but it's something , mine was overtrained for one pose and would not have this pose by default , tried few times)
https://civitai.com/posts/24153212
A bit slow but otherwise good lora.
Anyone knows how to merge loras properly tho ? :D Does expanding rank during merge help ?

@LocalOptima Lol. Right?! 1000 steps isn't enough for any concept that is not already solidly embedded in the base model.

@blo01 Once you merge multiple loras, the outputs for each individual part of the merge vs the original lora are bound to be different. Similar concepts and tokens will influence others and change the outputs.

And even if they did, that is all in the spirit of open source. No one on this site is making models from scratch, they just train them. And train them on data and imgs im sure none of them had hand drawn themselves. You follow the trail back and it all leads to someone eles efforts. We stand on the shoulders of giants. A better world is not one where we say "this is mine hands off!" but one were we say "look at this cool thing iv made, i hope you can enjoy it and run with it". It would be immoral to take something given to you for free slap some new lipstick on it then demand payment for it. Sure you can donate to people that deserve it like DigitalPastel here! But it they were to clam ownership over a Lora it would feel more like theft.

@fellowricky you're missing one key element here: The other creators are not farming the site's currency by deploying their LoRAs under Early Access. This makes your point evidently fallacious.

Besides that, the least you owe to your fellow creators is a mention when you've used their resources as the basis of your own. Try to use some common sense.

@meropewantsstls Early Access, not permanent ownership, I see nothing wrong with that. fallacious, my statement was not. If you were to credit every human that had a hand in a loras existence the list would be unrealistically long. Lets take the user K3nk for example as clearly they were the most irate. https://civitai.com/models/2061428/full-nelson-sex-i2v-wan22-video-lora-k3nk
i see 0 mention of any contributes to the lora. The researchers, the porn stars, nivida, none. Egos need to be tamed, this idea of being owed anything in regard to open source software is ludicrous.

@fellowricky I'm sorry. I was not aware of your cognitive limitations. Crediting people outside of the relevant environment (Civitai) makes absolutely no sense.

You mentioning Nvidia and researchers just proves that you cannot argue in good faith, albeit unintentionally. Using your failed logic an artist creating a statue shouldn't receive credit for their work because they didn't invent the chisel and hammer.

I hope this helps in some way shape or form but I doubt it will.

@meropewantsstls Your logic dictates that credit is only owed to the subject responsible by one step removed and only in a relevant environment. How often do you see the sculptor credit the chisel maker? Only one step removed and in the relevant environment. Never. Credit is not owed. Credit is given freely. A person took a tool and made something new of it. You can put what ever spin you want on it. it's the same thing.

Futher more, every single one of your statements are logical fallacies. Open with an ad hominin, then ambiguity, then some more ad hominin, slop on a straw man, followed by some more ad hominin. My cognitive abilities are more than adequate.

@fellowricky that's not my logic. It's basic human decency (as well as a professional and ethical standard) to credit someone when using their work as a basis for your own.

It acknowledges the time, effort, and creativity the original creator invested in their work. Furthermore (that's how you properly write that word, by the way), failing to provide credit is often seen as a form of intellectual dishonesty or plagiarism, which misrepresents the origin of the ideas or work.

In basically any academic, professional or creative fields, proper attribution is essential for maintaining credibility, fostering trust, and allowing others to trace the lineage of ideas and build upon them further. By failing to adhere to said standards, you're losing credibility and trust from your peers. This is only exacerbated by the fact that he is literally profiting from said work through intellectual dishonesty and lack of professional courtesy.

That's coz all the loras he merged there, you should start mixing loras and play with Lora str yourself.. I guess releasing a merged models wasn't enught, he went for the loras, so you keep this instead of the original Doggystylenlora from the original author, this is great my dude 🤦‍♂️

If you are using i2v, you can actually just use missionary lora with pov prompt for pov doggystyle.

every lightx2v loras can produce different results, for my facesplash lora the old lightx2v works the best

Well, you know you can rename that lora, right?

@LFAA Well, I can also slap a BMW symbol on a Fiat, but it won't turn into a BMW. Not sure what your point is. xD

You are missing the point, completely.
You are comparing chickens and prepared meal.

This is about a lora being created using OTHER lora:s. IE, someone taking chicken 1.0 and reshape that into chicken 2.0 without doing the heavy work needed to produce chicken 1.0.

This is bad manners, not illegal, but really bad manners.

The conflict has been resolved and a post has been made explaining it all and proper information about things used has been added and linked to in the description of this lora.
https://civitai.com/articles/22118

To be honest, you are questioning him without any basis and without any direct evidence. Are you trying to trap him in a self-proving fallacy through constant harassment and questioning?

Furthermore, this kind of behavior is unproductive and creates a hostile environment. Instead of making unsubstantiated accusations, it would be more constructive to present concrete evidence or engage in a respectful dialogue to understand the situation better.

@LFAA looking at the penis when the penis is not in the image to begin with is plenty evidence.. , anyone who tried training a good deepthroat lora will know.. I have a lot of versions, tests and reworks of my dataset,

A user that doesn't have a single deepthroat lora as a single model, I'm sorry , but I don't buy he made a better deepthroat action than us...

In civitai there's rules, you will be questioned about this if you don't add the resources list, and I can ask again if I see plain evidence... Seems like civit already took down one of the models.. guys , don't be a dick, respect others creators licenses and credit them..

@LFAA Dont disrespect the king/queen,

I personally don't care if people mix my LoRAs as long they credit me. I'm more annoyed by the fact that they're farming buzz while insulting the intelligence of other LoRA creators by pretending that training on 200+ videos for 5 epochs is going to produce the results this mix does.

Most of the LoRAs mixed in here didn't charge buzz when they were released so it's understandable people are annoyed that a mix that required a tiny fraction of the effort to make is charging for downloads. I add addendums to my model licenses that disallow people from selling my models or merges based off my models too, which means this merge is violating TOS.

"Criticisms driven my jealousy"

They're just authors asking for credit for their work. You can't be this detached from reality. He's literally violating the TOS on this site and this is what you garnered from the situation?

Ahem brother. All i see is a bunch of jealous egos yelling "i want whats mine!" Enjoy the art, share the art, and explore the experience that is life with love in your hart. Even if theses allegations are true, isnt it enough to know something that you have created has inspired other and brought joy into the world without the need to get your belly rubbed.

I'm actually using this XXX Lora on DaSiWa Radiantcrush and its pretty neat.

Clarify vaginal sex and in positive, repeat the action a few times in different ways and you should be good. I don't believe WAN image to video listens at all... not in my experience at least.

@Setian91 Prompt adherence is largely dependent on steps and sampler. In the former case, 12 steps for High pass is when it'll try to adapt the prompt more closely, broadly speaking. I2V source images might also skewer the prompt adherence if it's interpreting certain things.