ᴛʜɪꜱ ʟᴏʀᴀ ɪꜱ ᴘʀᴏᴠɪᴅᴇᴅ ᴀꜱ ɪꜱ ᴀɴᴅ ɪɴᴛᴇɴᴅᴇᴅ ꜰᴏʀ ʟᴀᴡꜰᴜʟ ᴀɴᴅ ʀᴇꜱᴘᴏɴꜱɪʙʟᴇ ᴜꜱᴇ.
ɪᴛ ᴀᴄᴛꜱ ᴀꜱ ᴀ ɢᴇɴᴇʀᴀʟ-ᴘᴜʀᴘᴏꜱᴇ ʀᴇꜰɪɴᴇᴍᴇɴᴛ ʟᴀʏᴇʀ, ɪɴꜰʟᴜᴇɴᴄɪɴɢ ʜᴏᴡ ᴛʜᴇ ʙᴀꜱᴇ ᴍᴏᴅᴇʟ ᴇxᴘʀᴇꜱꜱᴇꜱ ᴘʀᴏᴍᴘᴛꜱ ᴀɴᴅ ᴄᴏɴᴄᴇᴘᴛꜱ ʀᴀᴛʜᴇʀ ᴛʜᴀɴ ᴘᴇʀꜰᴏʀᴍɪɴɢ ᴅɪʀᴇᴄᴛ ᴛʀᴀɴꜱꜰᴏʀᴍᴀᴛɪᴏɴꜱ. ᴛʜᴇ ʀᴇꜱᴜʟᴛɪɴɢ ᴏᴜᴛᴘᴜᴛ ɪꜱ ᴀʟᴡᴀʏꜱ ᴅʀɪᴠᴇɴ ʙʏ ᴛʜᴇ ᴘʀᴏᴍᴘᴛ ᴀɴᴅ ᴛʜᴇ ᴄᴀᴘᴀʙɪʟɪᴛɪᴇꜱ ᴀʟʀᴇᴀᴅʏ ᴘʀᴇꜱᴇɴᴛ ɪɴ ᴛʜᴇ ᴜɴᴅᴇʀʟʏɪɴɢ ʙᴀꜱᴇ ᴍᴏᴅᴇʟ.
ᴛʜᴇ ʟᴏʀᴀ ᴅᴏᴇꜱ ɴᴏᴛ ʀᴇʟʏ ᴏɴ ᴘᴀɪʀᴇᴅ ɪɴᴘᴜᴛꜱ, ʙᴇꜰᴏʀᴇ/ᴀꜰᴛᴇʀ ᴇxᴀᴍᴘʟᴇꜱ, ᴏʀ ᴛʀᴀɴꜱꜰᴏʀᴍᴀᴛɪᴏɴ-ꜱᴘᴇᴄɪꜰɪᴄ ᴛʀᴀɪɴɪɴɢ. ɪᴛ ꜰᴏᴄᴜꜱᴇꜱ ᴏɴ ɪᴍᴘʀᴏᴠɪɴɢ ᴄᴏʜᴇʀᴇɴᴄᴇ, ᴅᴇᴛᴀɪʟ, ᴀɴᴅ ꜱᴛᴀʙɪʟɪᴛʏ ᴀᴄʀᴏꜱꜱ ɢᴇɴᴇʀᴀᴛɪᴏɴꜱ, ɪɴᴄʟᴜᴅɪɴɢ ᴍᴀᴛᴜʀᴇ-ᴛʜᴇᴍᴇᴅ ᴏᴜᴛᴘᴜᴛꜱ ᴀʟʀᴇᴀᴅʏ ꜱᴜᴘᴘᴏʀᴛᴇᴅ ʙʏ ᴛʜᴇ ʙᴀꜱᴇ ᴍᴏᴅᴇʟ.
ɴᴏ ꜱᴘᴇᴄɪᴀʟɪᴢᴇᴅ ᴅᴀᴛᴀꜱᴇᴛꜱ ᴛᴀʀɢᴇᴛɪɴɢ ɪᴅᴇɴᴛɪᴛʏ ᴇxᴘᴏꜱᴜʀᴇ ᴏʀ ᴄᴏɴᴛᴇɴᴛ ʀᴇᴄᴏɴꜱᴛʀᴜᴄᴛɪᴏɴ ᴡᴇʀᴇ ᴜꜱᴇᴅ ᴅᴜʀɪɴɢ ᴛʀᴀɪɴɪɴɢ. ᴀʟʟ ꜰᴀᴄᴇꜱ ᴀɴᴅ ᴘᴏᴛᴇɴᴛɪᴀʟʟʏ ɪᴅᴇɴᴛɪꜰɪᴀʙʟᴇ ꜰᴇᴀᴛᴜʀᴇꜱ ᴡᴇʀᴇ ᴍᴀꜱᴋᴇᴅ ᴏʀ ᴀɴᴏɴʏᴍɪᴢᴇᴅ ᴛʜʀᴏᴜɢʜᴏᴜᴛ ᴛʜᴇ ᴛʀᴀɪɴɪɴɢ ᴘʀᴏᴄᴇꜱꜱ.
ʙʏ ᴅᴏᴡɴʟᴏᴀᴅɪɴɢ ᴏʀ ᴜꜱɪɴɢ ᴛʜɪꜱ ʟᴏʀᴀ, ʏᴏᴜ ᴀᴄᴄᴇᴘᴛ ꜰᴜʟʟ ʀᴇꜱᴘᴏɴꜱɪʙɪʟɪᴛʏ ꜰᴏʀ ɪᴛꜱ ᴜꜱᴇ ᴀɴᴅ ᴀɢʀᴇᴇ ᴛᴏ ᴄᴏᴍᴘʟʏ ᴡɪᴛʜ ᴀʟʟ ᴀᴘᴘʟɪᴄᴀʙʟᴇ ʟᴀᴡꜱ ᴀɴᴅ ᴘʟᴀᴛꜰᴏʀᴍ ʀᴜʟᴇꜱ.
SideView Deepthroat - Minimax H3 LoRa - K3NK
Trained on Musubi tuner AkaneTendo25 fork with 119 videos and 42 flux images of my cumonface lora
Trained at 512,512
Rank 32
Trained video and audio but audio isnt that good (after more testing seems like audio wasnt trained in v0.1 , really cant tell since i changed the settings and didnt save the ones i've used)
All the example videos are made with 4 steps, still looking for the best turbo lora combo
Trained with Pruned model (will affect fluid dynamics)
Added v1.0 with a brand new dataset of quacking and gagging sounds, also added the raw gag5-lora , the training run with the quacking dataset, was lowres trained focused on audio, but might help to have the original lora separatedly as i'm using it along side this nodes)
The workflow im using is embeded inside the videos, drag 'n drop into comfyui
Example promptsFor the target video, at 0.00 seconds into the target video, <Picture 1> (from [Shot 1]) is fully referenced.
integrated_multimodal_description: [Shot 1] At 00:00.000, <Picture 1> shows the woman already in position, facing the viewer. She begins speaking, saying <d>[British English] "Look how this fucking bastard is throatfucking me" </d> with a calm expression. From 00:00.000 to 00:03.000, she continues speaking, maintaining eye contact with the camera. At 00:03.000, a man enters from [RIGHT] side, only his lower body visible, with a large erected penis. She pauses her speech and looks at the approaching penis. At 00:04.000, she begins to take the penis into her mouth, slowly pulling him in. The man has a powerful grip on her head sideways, hands in her hair, guiding her against his hips. She gradually accommodates the penetration, her eyes roll up slightly as she takes him deeper. The man maintains constant pressure, causing her to gag. From 00:05.000 to 00:07.000, the man thrusts his hips back and forth at a steady rhythm, she spits saliva occasionally. She moves slowly trying to release herself but he maintains control.
overall_soundscape: "Look how this fucking bastard is throatfucking me" with calm speech tone, wet slurping sounds as she takes him in slowly, continuous gagging and choking sounds, muffled gargling (GWA pattern) with each thrust, saliva spitting, man's rhythmic hip thrusting, faint hand friction. She CANNOT breathe or scream - all vocalizations muffled by cock. No screaming after initial phrase.
non_diegetic_music: None.No dialogue
For the target video, at 0.00 seconds into the target video, <Picture 1> (from [Shot 1]) is fully referenced.
integrated_multimodal_description: [Shot 1] At 00:00.000, <Picture 1> shows the woman already in position. a man enters from [LEFT] side, only his lower body visible, with a large erected penis. She begins to swallow the entire penis, shoving the penis into her mouth, slowly pulling him in. The man has a powerful grip on her head sideways, hands in her hair, guiding her against his hips. She gradually accommodates the penetration, her eyes roll up slightly as she takes him deeper. The man maintains constant pressure, causing her to gag. From 00:05.000 to 00:07.000, the man thrusts his hips back and forth at a steady rhythm, she spits saliva occasionally. She moves slowly trying to release herself but he maintains control.
overall_soundscape: Wet slurping sounds as she takes him in slowly, continuous gagging and choking sounds, muffled gargling (GWA pattern) with each thrust, saliva spitting, man's rhythmic hip thrusting, faint hand friction. She CANNOT breathe or scream - all vocalizations muffled by cock. No screaming after initial phrase.
non_diegetic_music: None.Have fun! 😜
Description
v1.0 Includes a brand new dataset of quacking and gagging sounds
FAQ
Comments (29)
Will these be sold elsewhere? I rather not mess with crypto.
Just wait 9 days, and get for free? Or is cock choking a time-sensitive thing?
The only way I can really seem to get good results is following your exact setup. I get that it's intended for I2V, but with almost every other lora I can get fairly normal results with REF2V. I also can't seem to use the 1.0 Lightx2v lora in any context, it seems to slow/mute motion greatly. I know I'm personally doing something wrong, but it's strange that most other loras don't seem to have the same issues when using slightly different turbo loras, or switching to REF2V. Any idea what's going on?
if only the throat bulge motion was possible, it would've been perfect.
It's only perfect if it can do male on male as well. But, all I see in the samples is women..
That said, throat bulge is entirely possible if it's trained using examples that show that.
@shanezwicker697 umm......
It's great but unlike the ltx lora it seems to struggle maintaining constant motion. For example if we want a no dialogue just pure action from men entering till the end -> a lot of pauses seem to be inserted? Is there a prompt example that you tried that doesn't have this? cheers
That's prompting not a lora issue.
use motion booster lora
Can you make a LoRA for face-slapping?
early acess bs, shame on you
very hard to add cum
Does dataset containing LTX generations? (version 0.1)
nope, i dont use synth datasets
What is the purpose of beginning the description with a bible anyway? Nobody reads this stuff, it's like installing a game you just click Next.
@FrogOnStilts Who said I was blaming anyone? It was a question. Don't try to put words in my mouth or imply some false narrative xD
We all get the ToS in our face when we create an account, we all know that nobody ever scroll through an entire ToS before installing or downloading a product, and it's already common sense that nothing should be used for illegal practice. So I just wanted to ask.
@FrogOnStilts That's a vague answer. So, why?
I don't generate videos of real people (including myself), I don't generate videos of minors, I never upload content, I don't redistribute files, I never used CivitAI's generation service, I haven't even had the time to use this specific Lora yet.
I don't need to read a ToS to understand common sense, so again... Why would I need to read the ToS?
@FrogOnStilts Uhh, OP does not need to worry about his account as long as he does not upload bad content, he has no obligation to repeat what CivitAI's ToS says.
But if the ToS says that content creators has to re-post CivitAI's ToS anywhere then I'd be happy to read it. Unfortunately neither GPT nor Grok could find such a statement in the ToS.
So what are you on about?
Me: Why repeat ToS?
Frog: Stop blaming OP!
Me: I don't blame. OP is safe, no need to repeat the ToS.
Frog: You should read the ToS.
Me: Where does the ToS say OP has to repeat ToS in his descriptions?
...Blocks me.
Ok so when faced with logic, yall just block the ones you're having a discussion with?
I JUST WANT TO SAY!! THAT THIS WORKS GREAT!!! FANTASTIC JOB! Expect some buzz to come your way!
It doesn't work at all.. tried it on Ref2VA and also FL2VA and i get the same grainy video. Same prompt as provided.
Which one provides best quality fp16 or fp32?
In the v1.0 i "fine merged" the loras so it creates more quacking audio, but still not perfect tho, I did this before I even knew about the audio refine loras, fp16 is the merged lora
@K3NK i believe i've got you not right, sorry. But If merged is quack, we love quack so fp16 is better, right?)
Tested v1 in several tests, differents models, included int8 and NSFW, and also NSFW lora,, or the lora alone+turbo lora. At 4, 6 and 8 steps. none of my tests, genrated a decent deeptroat action.
Lot of ghosting, artifact, and the main issue. Imposile to genrate a deepthroat scene, even a simple one, where the man insert his penis in thegirl mouth.
Before you ask or tell. Of course, i used the motion booster lora , itried several strengh, 0.5 to 1, none worked. and after 0.8 the motion is just ruined.
Another big issue with this deepthroat lora, the audio. The audio is horrible, even at 8 steps.
A lttle notice, i tested with T2V and I2V mode. As i do for all loras i firtly test. As the FL2V model provide a lot better quality video than R2V at the same steps.
For some reason, version 0.1 seems a bit better to me than version 1.0. This refers to animation.
yeah, v1.0 is a merge with the one that has quacking (you have the quacking lora as separated file) so i think at the end it messes up the action a bit, ive tried only merging audio but was hard, since H3 doesnt use separated layers..