Also the LTX-2 license states that you cannot sell anything trained on its base model. So they are in violation :p

@Malektha Sounds like this site may be in violation of that license by allowing users to require a form of payment in their "early access" program...

yea i cant see a good future for these people. they are going to learn the very hard and very expensive way.

@playtime_ai_ valid point, might be some legal loophole in regards to converted currency tho :s

@Malektha True... There probably is some legal loophole created for billionaires to extract money from plebs that can be used for such purposes. I don't doubt it.

@playtime_ai_ I think only in the cases where the thing doesn't leave EA and is only purchasable is where it gets greyer, ALA biglove.

@happylittleteapot The smart creators - FLUX and everyone, are doing it so they can say in court, "Well, we told them not to sell it!" when they get sued by the people they stole the artwork or photos from, so there aren't damages. The greedy creators don't get that's why people do that and think they can sell licenses.

I'm not saying their case would hold up in court, but these guys actually do put in time and skill into their work, it's a real product they are offering. But on that note, I personally keep using other people's songs from suno, and would be very happy to watch the situation unfold with a front row seat if one of them tried to sue me lol.

much love! thnx for the positive feedback!

Agreed!

commenting to follow thread

he is selling a product.. i dont think he will share..

@K3NK defo worth asking haha , but aye

@K3NK so bad because a lot of us can provide really good stuff, and différents ideas . It's seems they trained their lora on each aspect, audio is very clear , i suppose they have a specific dataset with pure audio, on this one there is not full penis, it helps a lot, motion as i tested in my loras is easy to train , it's very fluid , train with 30fps i suppose.

@K3NK Open sourcing the method and technique would still benefit their business. Imagine if we're gonna see a flood of really good community Loras. They can all use it for their commercial products without paying for labour and compute to compile dataset and train it.

@RogueStudioAi why only telegram? do you have a discord instead?

I'm seeing the same thing. I'm playing around with different configs to see if I can get something more petite.

lol - yeah it really is best for image to video -- and agree the breast are too large for t2v -- will release a v2 soon

@hdfsdfsfsf600 Thank you! We're complaining about big breasts, but this lora really well done!

i2v give boobs too

I do not have that problem, i can prompt the breast size. Even if i do not prompt it i get normal breast sizes. Thanks for the great lora

@gambikules858 Not for me. You might have the strength too high. Read the description. 0.5 is recommended.

@LetTheBassDrop I've tried with a bunch of weights (and a few other minor tweaks to the workflow) and all result in large breasts. I'm using the standard comfyui ltx 2.3 workflow, with 2.3 weights. Would you mind sharing what you're using?

@AndyZocker What are you using to generate? 

@sugarfreewafers Distilled model and lora on 0.80. But using another workflow with live preview built in, not sure from where i did get it. Breasts are not tiny or huge but i would say in the middle most of the time.

@AndyZocker Thanks for sharing! I'm using dev model with distilled lora. I've tried a variety of weights for both distilled lora & rogue-missionary-cowgirl lora (0.5-1.0). Also messed with cfg guider and steps. I guess size is subjective, but definitely haven't been able to achieve the size in the examples provided (sounds like those are i2v tho). I'll try just the distilled model and this r-m-c lora and see if I can get smaller sizes.

It's not about your specs, its about your knowledge of the model and almost certainly your prompting. Working with these models is not a set and forget deal where you instantly get amazing results. It requires knowledge, trial and error, iteration. Read up or watch tutorials friend, and instead of downloading someone elses workflow and expecting it to magically work, LEARN what each of the nodes is doing and why. Even better, get enough knowledge to build your own workflows.

save one of the examples videos and drag it into comfy.

are you using the Ltx2.3 i2v workflow? where do I ad a lora to it, trying to use the RogueAI one and it comes with no wkflow, and the Ltx2.3 wkflow has no powerloaderlora node? How can i get this to work?

@irishpc491 every workflow you will find on another lora here will work , just search a bit more :)

same issue here