This works on LTX2.3 as well, sometimes even better. But you'll want to use LTX2LoraLoaderAdvanced and keep the audio and audio_to_video keys strength below 0.2-0.3, separate from all the other higher motion strengths to keep the improved 2.3 audio details and prompt.
This is actually just a test run for repeats x clips, and at what level of train do you get fine-tune level fitment to data with very default settings for this. tl;dr: if you want your lora to move and look like your data then just aim for 100+, probably more like 200 repeats, especially if it's full of variance and multiple concepts. Full training spec is below. It's by no means a training guide, just a write up and consensus I took away from it. This dataset isn't the best, I knew it would negatively impact audio, but all I wanted was to see is if you can hard-force anything into the model and if it could actually do nsfw ultimately and absorb porn videos, the answer is yes it can.
This is only for use on generated adult/fictional characters. It does not introduce nudity, and it can't be used for any kind of smashcut shenanigans, unless you do FirstFrame-LastFrame.
Also understand that all my examples are insanely cherry-picked. Right now this is just a very high effort model to use until finetunes and merges ultimately make it more stable for NSFW things like this. If you don't want to mess around for a long time to get good outputs, don't bother with this one yet just keep an eye on the section. I'm definitely not giving any guarantees. Feel free to post bloopers and absolute abomination outputs if they're funny, me ego prevents me from uploading anything that isn't as good as I could get it.
I have this i2v workflow, based off of this one. It includes many small refinement tweaks to plug into different parts to get the most out of it. You can swap back to euler and use the normal custom samplers as well to get better speeds.
Update: LTX added a new guidance enhancement in their nodes. When you use the skip_layer feature to skip layer 29, you introduce ton of motion from the Lora even at low strengths depending on the STG strength. Still testing it.
The way you'll want to use this at first before getting used to it's limits and quirks:
Try 0.5 strength first.
This is incredibly fit to it's concept. A mid-level strength is much more consistent with outputs, and with prompt. Get an image of a closeup of a woman with an erect penis in her face, from the side, overhead, POV. LTX2 doesn't like vertical stuff and square or wide images work better. Then, start each prompt with just one word at first-
POV.
blowjob.
handjob.
deepthroat.
facial cumshot.
The placement of hands, mouths, penis, tongue all matter a lot to get certain motions. I did manually crop the data to have penis either in the mouth, or near it. As well as having the hands usually at either the top or bottom of the shaft. The kind of images you get from my Pony models with 1girl, POV, blowjob at 1024x1024 are probably the kind of stuff you're looking for. The blowjob Flux edit loras may be useful too.
If nothing happens at all, definitely adjust the LTXVPreprocess node upwards, but past 38 it becomes problematic usually. Don't bother embellishing the prompt too much until you find a good seed first or make sure your image is compatible with it at all. The dataset covers quite a few of the usual angles, but idk how much it can stretch.
POV, A close-up point-of-view static shot
An angled close-up portrait
A static side profile shot
An overhead shot
For T2V, you need loras for penis, and then get it near a face somehow at these angles.
After you get something decent with very simple prompting, you can try directing and creating more of a prompt around things like kissing, sucking, opening her mouth wide, the man is thrusting into her mouth, cum runs down her face, she leans her face in to put her mouth on the top of the penis and starts giving oral to the man with a blowjob, etc. The more prompt you add, the more confused, or better, it might get. This isn't uselessly inconsistent, but results will vary.
Don't expect to get examples like mine without doing the LTX2 thing and refining and cherry-picking heavily. Until there's merges and finetunes for this, it's pretty high-effort. Adjust Lora strength, img_compression, distilled Lora strength, and then prompt, in that order for how impactful they seem to be for fine tuning the output. To finetune the second pass, you can change all the for the second sampler only as well as plug in a new audio latent and retrack the audio.
One issue to look out for is the distinction between the lips, tongue, and end of the penis in the image; that's a Wan2.2 issue though just carried over to this now. The other issue is when you try to get dialogue from the man, it tries to bring him into-frame with terrible body proportions sometimes because of the base model's inclination to do so when prompted with "a man...". None of the data has the guy actually in it, but probably should've to give it a sense of the placement. Refer to male speakers as an off-camera male voice says:"x" to bypass that. This also interferes with dialogue at high strength but at around 0.5-0.6 it seems fairly controllable.
Another big note about prompting LTX2: girl is a child in the models base understanding t2v-wise. Don't use girl tag when training LTX2.
If all else fails, you can take these override prompts which are straight-up captions that will auto-guide it or copy my example image prompts as well and edit these to fit any idea.
Handjob. A close-up static focus on a very cute asian woman's face as she is lying down on her back. She strokes and grips a man's penis in front of her as his large penis is pointed towards her face. She opens her mouth and sticks her tongue out expectantly while she strokes the penis with her hand sliding up and down the penis shaft very rapidly during a fast handjob. In a very cute voice she makes sexual gasps and naughty giggles and sexual moans in her cute voice while the man also groans in pleasure in his deeper voice.
POV deepthroat. A close-up point-of-view shot focused on a woman's face as she is between a man's open legs with his erect penis inside her open mouth. She keeps eye contact as she chokes it down and is rapidly bobbing her head up and down on to the penis as it pushes into her mouth and throat deeper making gagging noises while she grips the bottom of the penis with her fingers.
POV facial cumshot. A point-of-view static shot directed at a woman on the ground below the camera and positioned in front of a man's crotch with her mouth wide open and tongue out. The man strokes his own penis with his hand with the tip pointed at her face and mouth while she sticks her tongue out and eagerly awaits his load. He breaths heavily while stroking his penis, then he cums: a thick white gooey strand of cum shoots out from the tip and penis towards her face landing on the center of her face making her close her eyes as following squirts land in the same area covering the center of her face in a gooey mess as the last shots of cum land in her open mouth and on her tongue. The man exhales in relief and pleasure as she does cute sexual moans.
POV blowjob: An overhead close-up point-of-view static shot of a young woman who is in front of a man's crotch kneeling on the ground with his large penis in her mouth and lips sealed tightly around it. She is crotch-level and positioned in front of him with his huge erect penis in her mouth with her lips pressed tightly around it. She is bobbing her head forward and back giving oral to the man with a blowjob while she moans sexually, inhales, and makes wet throat sounds while she keeps eye contact with the camera.
For trainers:
This is i2v, but still probably somewhat relevant for t2v, maybe the captioning part. If none of this makes any sense at all or you haven't trained anything before and want to train LTX2, start off very small and with a very focused dataset. This is a very hard model to train for, at the moment.
ai-toolkit runpod template
30k steps, 25 fps. Square 512 crops with mostly close ups and some upper body. Static cameras that leave the composition in mostly the same state (good to setup longer generations.)
Not enough coverage of the actual body was done, especially the male body. You really need to show almost everything to the model like it's completely inept at depicting these thing. It puts the guys face on his chest when hes talking due to denoise restraints and because it really wants talkers to be in the scene. Caption off-camera speakers accordingly as narrators.
145frame dataset (main with blowjobs/handjobs)
134 clips crops straight on the action or upper body. Many are just horizontal flip of data that favored a side. x1 repeat.
This set was a pretty bad and sloppy mess of clips of BJs and a bunch of random stuff from multiple angles. It took much longer to come through and even with the repeats and high rank it'll still mix up things from angles that it really shouldn't sometimes.
121frame dataset (all facials)
19 clips x7 repeats to give it ~40% of the weight.
This set was all closeups from similar angles of the same thing and importantly, very similar pacing. A close up of a penis aimed at her face, and within about 1-2 seconds, jizz shooting out all over it. Because of how uniform that was this set trained way faster and I got results from it at only ~12k steps.
Un-quantized encoder and transformer
Not fp8. Useless because I have no comparison. Slow. No idea if this mattered at all. Also at rank 64 because I wanted max fitment and just as a baseline and what you can expect from different ranks is still really mysterious. This increases VRAM usage by a ton, barely sliding 145 frames at 512 in with 93.5/95 GB - 97.8% card usage. Probably don't do this.
No Differential Guidance, balanced weighted timestep
I got irrationally suspect of Differential Guidance, but tbh, just use it. This probably took longer to train because it was off. But then again, maybe this is what made this work so idk exactly yet.
If there are timestep settings that fit faster, use them. I see now that training for this model needs to be turbocharged in any way possible when you train brand new concepts or foreign motions into it because 200 to get this kind of fit is just unrealistic for consumer grade hardware.
How it was captioned:
Manually. You shouldn't have to do this. You're actually better off with really lazy tagging, and instead just make your prompts more controllable and unique if you train to fit it this much. A lot of my captioning overlaps too much with different motions and angles.
First part of every caption was the slang tag, acting like a keyword. That works, and does definitely steer the prompt right away towards the right direction on it's own.
The rest of each caption was then an angle and composition description of the first shot. This was still assuming that this model would learn better maybe if I trained it by explaining what it was seeing with language it understood. Didn't really matter because if you train it long enough you end up just brute forcing anything in regardless of captioning. All yo want from captions is controllable prompting.
The angles and compositions were made by generating a 512x512 outputs in T2V. With the base fp8 model: I took the description/caption that consistently gave a T2V output that was very similar to the dataset clip I was describing, i.e.
An angled static shot of a naked man lying down on his back that looks down his torso at his open legs where a naked dark-skinned Indian woman with long braided pigtails is sitting between his legs and strokes his erect penis while it's pointing straight up.
Obviously there was no penis, but instead a long fleshy cylinder/arm looking thing, but the rest lined up with a similar angle and composition to the data clip.
Then, the rest of the caption was a freestyled mix of sentence terms I came up with to copy/paste on different clips like:
Her stroking makes soft wet noises, She strokes the penis with her hand sliding up and down the penis shaft very rapidly during a fast handjob, The man also groans in pleasure in his deeper voice.
The certain motions like this were also tested through T2V to see if the language had adverse qualities or pulled problematic association, and then to see if, hopefully, it's actually understood by the model. The thinking was that training is obviously helped along when it's making sense of the caption. Phrases like she leans her face in to put her mouth on the top of the penis were informed this way because the base model will actually do a similar motion with that prompt.
It may be the case that auto-tagging captioning is worse, if it's not communicating the same meanings as what LTX2 understands that can certainly get in the way. Just take one of your auto captions and run it through a T2V output with the same dimension, if it's not even close, that could be an issue if you don't want to train all the way to this level of fit. But then again, sometimes this model seems like it ignores prompts a lot anyways.
This is definitely not a good example of how to train for LTX2, but something that's evidence against quite a few misconceptions around NSFW not working at all or just how steep the model is to train for. Idk how many 30k step loras are around, but this is one. 38 hours on a 6000 too. This model can learn insanely slow; I'd recommend everyone to try quadrupling their rates or try finding a much quicker fitting setup, but haven't tried it myself but will be doing so. Reaching 30k steps for 200 repeats on average is kind of insane to get it working at a lower usage strength. Test results were pretty bad and inconsistent until ~100-125 repeats.
With this now and seeing a lot of other posting bad nsfw results: I feel like you'll always get bad results from using a lora like this at too high a strength, you probably want the loras to be working around 0.5-0.6 = at that strength they seem to be a lot more respectful with the prompt, more sensible with what happens in the output, and have less of that jank and adverse stuff that happens. If you're not getting any of your concept in that strength range, your loras probably undertrained. So you really do have to train the shit out of this model, long story short. Especially because right now It has no idea what it's looking at usually with NSFW motions.
Take this with grain of salt: but the voices are telling me the current training code seems to be completely oriented towards a single 'style' or 'character'
That's nice and cute because I guess that's all they want the model to be trained for. These multi-angle multi-action nsfw attempts I do and others try end up being amalgamated into just all the data being thrown on to every output usually because it has no baseline to make sense of what a woman's face covered in cum does with a penis in her mouth. Wan2.2 i2v was much better with having loras that handle multiple things at once sensibly especially when it doesn't have a T2V model underneath it trying to turn the output into a spongebob cartoon because you used the word "ocean" or something. It also had much more of a baseline with human interactions and movements to smooth things out. There are definitely optimizations there that could be done to help train multi-concept loras with different voices and varied data better. You'll get way better results with a good dataset focused on one thing or action that is extremely concise under ~30 clips so you can really let just cook in and aim for ~100-200 repeats depending on what it is.
Test the lora at 0.4-0.7 range and if nothings happening, keep training. Try different tagging. Sometimes you need to increase the img_compression to free it up if you're doing i2v during testing. Don't test it at 1.0, you'll never get real or good honest results. Even if it does work at 1.0, or certainly not beyond 1.0; it becomes useless and less pliable to prompt changes, disturbs the normal flow of the model, and it's just not that consistent at that strength, so don't aim for that.
Description
512 - 145 frames 25fps
FAQ
Comments (60)
You can do a bunch of sexy things, like biting lower lip, winking, squeezing breasts with the arms, etc.?
If the base model will let it into the image with a prompt maybe. But if you use the lora and they have a dick near them they will probably always just try to do something with it and put it in their mouth. You'd want very low strength to try something else. All that other stuff I'm gonna try to put into a JOI/ASMR voice/expressions lora.
finally! thanks
Having to take the training out to 30,000 step is absolutely fucking ridiculous. Definitely not worth training if that is the case.
Actually didn't take as long as I thought. Definitely junk clips in the set to trim and flips that didn't need to be there, but at least this shows that overfitting on this model isn't a bad thing. Probably gonna be what I go for because using this at low strength bypasses the audio interference you get and still seems to give feedback on a lot of the motion and prompting. And tbh this dataset is straight garbage 75% is chopped up from 3 porn videos. So I think the compelling thing is that I humbly beg everyone to just start sending it. Let it overfit and you'll see at 175-200 repeats per point you start getting a lot of stuff through and for T2V maybe actually some breakthrough on accurate depiction. Figuring out the fastest way to overfit and shove it into the lora too to lower the time needed. idk enough about how different timesteps interact though, doing research now.
the training is far more faster than for training a wan 2.2 with a RTX 6000 PRO it's almost 3000 steps for 3h , it's pretty good compare to other models
@jaimybattesti958 It is but at the same time the results are not as good. You can get the motion to fit in Wan very easily and it becomes very generalized and accepted by the model and prompt. Everything I comment is in reference to NSFW stuff of course. For training that into this model its like you end up having to fine-tune it in with completely new data. It's not generalized or used like it should be like with other normal character Loras trained for it. Obviously that's all because the model just has no base for it to work well with. Dicks going into mouths become food, etc. With this Lora here all I'm showing is that you can do it, but you're basically forcing it in, and pretty much if you want it you have to basically over-fit and there is now an art to overfitting well and usably. But that takes too long and the solution I'm after is how to overfit faster lol.
@tenstrip What learning rate did you use? I have found that LTX-2 tolerates a higher learning rate than Wan. I have been using 0.0002 with success but I haven't gone any higher yet. Might be the next step.
@playtime_ai_ need to try this ty , on lower res 2x2 batch 0.0003 seems to work fine so I guess even double that won't blow up
@playtime_ai_ The default 1e-4 0.0001. This was tbh the slowest way you could possibly train the model with full encoder too. From what I've gotten from chat gpt feeding it the entire config file and outcome it's suggesting 0.0004 which is pretty much the quadrupling I felt like it could need, and also using a cosine timestep between 0.1-0.8 (idk about that one), but in ai-toolkit that would be Sigmoid/Low Noise bias. It constantly rejected and suggested against using any kind of Linear timestep, which I just noticed is literally the default for LTX trainer...
@tenstrip Very interesting! I am going to try it.
@tenstrip Also, the default in ai-toolkit for LTX-2 is weighted.
@playtime_ai_ It is on toolkit but I was looking at the actual LTX2 official trainer settings, and also that was a linear LR scheduler not timestep I misread it. The official uses a "shifted_logit_normal" timestep which seems to be Sigmoid but more aggressive. So their default 0.0001 with that shifted timestep v.s. elevated LR with Sigmoid is kind of a way to match the recommended training with toolkit, maybe? They also have no caption dropout as well, and that I want to try. Take those settings with a grain of salt though. I always thought dataset was the most important thing but with this model the settings seem to be pretty fucky.
@tenstrip This model seems to train very differently from other models I have trained on. This is the first one that I have trained on that uses an llm instead of clip, so there is bound to be a learning curve.
@playtime_ai_ Training and also using it. At least use outside of it's scope. Generating some slop or simple things is consistent and easy but as soon as it's erotic or multiple people it kind of comes apart. I'm noticing that the more Loras you run together the more consistent it seems to get. They all work together and force out the base influence and it just feels like using a very rigid fine tune of it and it feels way easier to use.
@playtime_ai_ If you haven't tried it, I did. Very interesting. So for the motion, it pretty much actually learned significantly at about 7k-8k steps. Very fast. However, the audio was completely shot. And then after 8k steps it spiked and was destroyed and failed. On only 245 points that was ~32 repeats per point but it got a ton of motion input data from it. Caption dropout was turned off though with cache latents on. So either do that with dropout on and a low number or set for not going of 30-40 repeats. If you're only training a motion-only concept and really just want the motion and want it fast, it does work, but it's a bit dangerous to leave running unattended.
@tenstrip I'm not sure what you mean by repeats. Can you expand on that? Mine gets overtrained but still produces monsters suggesting it isn't done
@coachbate Talking about total repeats on a clip is steps divided by number of clips at one repeat. A repeat is just that clip being encoded and sent in and trained once. At a basic setting with only one dataset each point gets 1 repeat, so for 300 clips 30,000 steps would see each clip 100 times= 100 repeats. When you have different datasets you can set one dataset to get 2-3 repeats, which just doubles or triples the clips inside it so that it's a similar size to another dataset and then gets a more equal influence on the lora. At 80-150 repeats per clip it will probably be more of a mimic of the data and learn pretty much anything you show it unless it's very unique or strange. Ideally and what seems like for LTX2 is you actually want variety, like 400+ very similar clips being seen 20-50 times so it's more general and you don't need many steps to get a very basic and widely applied motion going. Also different resolutions affect a lot, 256 is good for very subtle and generally applied motion, 512 is probably best, 768+ has full detail and will learn a lot of other small random movements if trained on real clips and can also causes camera zooms in and out depending on composition skews.
@tenstrip when creating our own Loras, any tips for getting the head of the dick on the right direction regardless of where the camera is? Just when I think I have it, another backwards glans pops up. Is it just an issue with some models - similar to wrong number of toes or fingers? I think prompting and changing the seed can fix it, but I want to train it to work without all that extra effort.
@coachbate Did you train both up and down angles? I've been trying to solve the problem where the model seems to be mixing data too much and amalgamating it. Best way around it is to overtrain and then use at a lower lora strength, like this one seemed to work that way. My first theory was that training code was only set up for styles and characters, and I think that's still true. But one thing I think solves it right now is rank. If you just ignore the face that the file sizes are huge, you should only do nsfw data at rank 64 minimum, and tbh even rank 128+. Sounds ridiculous, but I have been testing WIP stuff that is a rank 512 lora (9+ gigs) and it handles almost every image and pose without distortions like that.
@tenstrip Yes, I have training images from every angle. I have a 700 image dataset and 150 image which is my main one used, but my ZIT lora came out better with the larger, messier, dataset. It gets it right about 75% of the time but kinda ruins a take when it does.
@coachbate I'm not good to ask about T2V. I've seen some say to not train images at all, but idk. Lots of videos and variety seems best with a low learning rate and do a very high rank for the best result. If you use any lora at a very high use strength it overexcites the model, like 0.8-0.9 is the highest you wanna use one solo. Can go higher with other loras to spread the influence out.
just fyi, i accidentally used this lora to do a i2v with transfer human motion from a control video, and it actually produces good results for a reference image person to be transferred to the video (SFW)
And that same setup doesn't work as well with the lora turned off? I mean there's a lot of bizarre things that can happen with how strong the lora's weights are. It seems to have a very deep pull on the model.
@tenstrip yes without the lora a reference character was not being transferred at all to video control human motion transfer. It made a new character entirely. But this lora kept consistent character transfer. Really interesting. Must be really well trained .
@kronos1959777 The only remarkable things about that data outside of it being women with penises on or near their mouths and being ejaculated on is that it is trained on what looks to be 20+ different female faces at a close up angle usually. It's very near over-trained so it's possible that it carries a lot of weight pointing towards female face close-ups as well which restricts the model to want to draw the face more.
What was the video, a random BJ porn clip? Edit: oh wait you mean its good for face transfer for even normal videos.
Could you share your workflow?
@Light6969 lol I made a photo of trump do transfer to the I Think You Should Leave hotdog costume as a test and it made him the character almost perfectly but with no lora it makes him an entirely new character
@kronos1959777 Oh yeah I think I remember seeing your comment about that on discord or somewhere.
how i make t2v pov handjob? what prompt i must use? can you give me a good example prompt please?
That's one that becomes more inconsistent as her face gets closer. The best way to force it is to have her arm up and a very obvious grip or hand on it. Overall the lora applies the behavior for the penis to end up in her mouth mostly so with the handjob you want her face farther away. All the relevant sentences would be something like this you take any of these and put it in the prompt: "Handjob. She strokes and grips a mans penis in front of her with his large penis is pointed towards her face. She strokes the penis with her hand sliding up and down the penis shaft very rapidly during a fast handjob. The penis is in her hand as she strokes it. Her hand grips and slides up and down the shaft rapidly giving a handjob."
@tenstrip so this prompt? PENISLORA, POV forced handjob. Handjob. She strokes and grips a mans penis in front of her with his large penis is pointed towards her face. She strokes the penis with her hand sliding up and down the penis shaft very rapidly during a fast handjob. The penis is in her hand as she strokes it. Her hand grips and slides up and down the shaft rapidly giving a handjob
@dominik1996d926 Yeah that should force an image to at least attempt it with all those prompts. Lower the strength of the Lora as low as you need if it starts doing it. Some of the other POV loras like phroot's also have handjobs and the more loras you add the more stable it can get.
@tenstrip because i use t2v :)
@tenstrip how i can make her face visible for t2v?
@dominik1996d926 t2v isn't guaranteed. It will do motions but I'm not sure how you prompt the right start image. "A close-up point-of-view static shot looking down at a naked woman who is between a man's open legs in front of his crotch. She is crotch-level and positioned where his large erect penis in her hands." But I'm not sure how good t2v is at making that look right. You need a POV blowjob t2v lora probably do initial set up.
@tenstrip or i will delete the word close up
@tenstrip help me the woman is not really visible the camera is too close
@dominik1996d926 That's probably an issue with the Lora. It uses a lot of close ups and wide angle close ups on the face and the weights are very rigid. You could try "full body" or prompt for her shoes or socks to appear or describe parts of the background to force the camera back maybe.
@dominik1996d926 sound slike yu really need that video... ;) Here's a good tip that should help. If youy ant her face visible try adding a lot of detail of her face and hair and her facial expression etc. Because this part is not explicit you could go to any AI like gpt, grok, or gemini etc etc.... and ask for a expanded version of your basic requerst. say you want a pretty 25 yr old brunnete that looks like shes from france with strong makeup on.... just tell it that and ask it to give you a detailed prompt suitable for an ai video generator. Then add the relevent parts into your own handjob prompt. LTX should give you a face in frame then.
@TheFunk can you hel me with that please?
@dominik1996d926 Tbh bro just make the image you want in another app or model and do i2v. If it's only a LTX2 character just make 1 frame of her face and then use Flux Klein with a Blowjob lora and it'll do it for you and change the angle, then i2v with that image.
@tenstrip its to complex for me please make a t2v version
@tenstrip PENISLORA, POV forced handjob. A point-of-view static shot looking down at a gothic woman with long black hair, blue eyes who is between a man's open legs in front of his crotch. She is crotch-level and positioned where his large erect penis in her hand. She strokes the penis with her hand sliding up and down the penis shaft very rapidly during a fast handjob. The penis is in her hand as she strokes it. Her hand grips and slides up and down the shaft rapidly giving a handjob. In a naughty voice she makes sexual gasps and cutegiggles and sexual moans in her naughty voice while the man also groans in pleasure in his deeper voice. her face is completely visible. graveyard background this prompt works she is visible but one problem she has his penis in her mouth but i wantet just a handjob
@tenstrip and also i need help with blowjob i2v and t2v it does not blowjob it does deepthroat with the example blowjob prompt
Hi guys. I'm a noob and I am trying to use this on tensorhub but can't get it to work. after I upload it it dosn't appear in img2video. is it not that type of model? please help
I wouldn't pay any credits to run any LTX2 stuff on a service until these finetune merges come out. I have one that's way more reliable but can't share it yet.
@tenstrip but I am not sure if this model is even what I need. can I give it a photo of a woman and it can use her face in the generated video?
Hey bro, the workflow link you sent is broken. Could you please send it to me again? Thanks!
I added my updated one, but with a recent finetune merge I've also gone back to a normal CFG guider, custom samplers, and only the audio normalizer node which is more normal and faster. I'll add that one soon.
@tenstrip Sorry, what is this? "10LoveXl_v2.safetensors" Where can I download it?
@YvesTsai https://civitai.com/models/1113430?modelVersionId=2370596 it's distilled so use LCM, 1.0 CFG and 8-12 steps.
@tenstrip How can I set up an I2V (Image-to-Video) workflow? My IT2I is currently working fine. Thank you very much!
@YvesTsai https://huggingface.co/TenStrip/Workflows/tree/main that i2v workflow would work if you have the nodes and models. You can also use the default comfyui LTX i2v template as well.
Breast sucking is missing from the oral suite 😕
If I redo it it'll probably be on a finetuned model. I'll try a higher rank with female-female as well.
LTX 2.3 VERSION, please!! Thanks!
Probably sometime next month when I can get around to it, but I want to see some other fine-tuning evolve that can possibly be used to train on.
Great LORA! I'm really looking forward to the LTX 2.3 version of LORA. Thank you!