My second lora is ready :D
This is the "pov" of a 69 position blowjob deepthroat / throatpie.
⚠️WARNING: the following LoRa might produce unwanted tattoos, and earrings, user discretion is advised 🔞 😂
i tried to caption those but some how they will spawn anyway..😥
a closeup pov view of a 20-years-old woman in top of a man with his penis inside her throat, she has small earrings, she has a tatoo on her arm, the man has tattoos on his arms, she is on top of the man sucking his cock, he grabs her by the neck and pushes her against his balls, bursting his penis inside her throat, povviewdeepthroat, noseonballs, faceslam, headgrab, throatfucka closeup pov view of a brunette woman in top of a man with his penis inside her throat, she has her head on top of the man penis with her mouth open as he burst his penis inside her throat in fast up and down movements, the balls of the man bounces up and down, he holds her neck with his hand, she is looking at the viewer, povviewdeepthroat, headgrab, throatfuckit also contains a couple of sideview clips on the dataset
a closeup sideview of a woman in top of a man with his penis inside her throat, she has hoop earrings, she wears a leather harness, the man has the whole penis inside her throat, she mantains the penis inside her throat while the man grabs her by the hair, she takes her head up and pulls out the cock from her throat, strands of drool and saliva comes from her mouth, sideviewdeepthroat, penismantain, penisrelease***** This LoRA is shared without warranties and with the condition that it is used in a lawful and responsible way. I do not support or take responsibility for illegal, harmful, or harassing uses. By downloading or using it, you accept that you are solely responsible for how it is used. *****
Description
This is the same dataset but captioned with JoyCaption for more accurate captions, from the epochs i tested , up to 78, this was the one that was more accurate with the prompts.
FAQ
Comments (23)
You're the MVP for including full workflows.
Any idea why It's extremely satured when I do it?
The most complete LoRA. I'm looking forward to your other works as well.
Could I get a workflow for this please?
have you checked metadata? some told me it wont load, the workfow im using is the allinone ultra 1.3, youc can also try any other workflows from @LatentDream https://civitai.com/models/1007385
@K3NK I just opened up that workflow and it looks so overwhelming, maybe video making isnt quite for me... yet ;P
@K3NK please ignore me i downloaded the ultra versions lmfao
Amazing, I would love to see a pony lora for images.
Amazing. Do you have a link to the guide you followed to make this lora ?
i installed the wsl image https://civitai.com/models/1085714?modelVersionId=1219185 ( i have v1, no webui, and it has the models inside the image) ,
then you can follow this guide once you have it installed https://civitai.com/articles/9798,
and finally for the .toml files you can find them here https://civitai.com/articles/10686/hunyuan-video-configs
As for the dataset, I'm starting to realize, after no one has pointed it out, that you don't need to trim and resize all the videos, just set the resolution to something like 211 in dataset.toml and you're done... just don't use very long videos; all I do now with the videos in my dataset is crom the video to remove any text, logos or parts I don't want and trim them to a maximum of 3 seconds per clip.
Good luck!
@K3NK Thank you so much. A couple more questions:
What GPU do you have ? I have only a 3090 hope thats enough.
How many video clips did you use ? Were they very high quality ? I know high quality is the most important parameter for images, but not sure for video as it seems like it get resized to a smaller size anyways.
Thank you so much for sharing. I cant wait to try to make my owns.
@Ggrue i have 4090, but i think you can train with the 3090, just dont set more than 211 res in dataset.toml
as for the resolution of the dataset videos, i think it doesnt matter, just dont use long videos , and try to keep ip under 1080p for the dataset material, i think youre good like that..
Can you please make this for WAN?
This is a very good LoRA. How did you manage to exclude the face of the training dataset tho?
just mixing various faces, the end result will be a unique one
cool，Is setting the latent window size to 9 acceptable?
When I use FramePack F1 to generate a 5-second video, I notice that for the first 4 seconds, a woman moves her mouth close to a cock. In the final second, she only lightly holds the tip of the cock with her mouth—there's no bobbing motion before the video abruptly ends. However, the reference videos I've seen all show 5 seconds of her holding the cock while actively bobbing her head. Could this be caused by an incomplete LoRA action due to the latent window size, or is it a seed-related issue? This problem has been troubling me for a while😭
Prompt: [0] a closeup pov view of a woman in black maid dress, short black hair and red eyes [1] she is on top of the man moving fast up and down his cock, he grabs her by the neck and pushes her against his balls, bursting his penis inside her throat, indoors, white background, The camera remains fixed throughout the scene, povviewdeepthroat, noseonballs, faceslam, headgrab, throatfuck, cum in mouth, cumming,
Negative Prompt:
Seed: 1306280934
TeaCache: Enabled (Threshold: 0.15)
Video Length (seconds): 5
FPS: 30
Latent Window Size: 9
Steps: 25
CFG Scale: 1
Distilled CFG Scale: 10
Guidance Rescale: 0
Resolution: 480
Generation Time: 1 hour 47 min 32 seconds
Total Seconds: 6452 seconds
Start Frame Provided: True
End Frame Provided: False
Timestamped Prompts Used: False
LoRA: 69Deepthr0at-throatpie-v1.1
LoRA Scale: 1
the sampler were made with the allinone workflow, t2v, i actually havnt tested that much this lora with framepack, try using 15 guideance_Scale , latent window 9 give me better results un cumshots for example
@K3NK LoRA is effective, thank you. Perhaps I should also change the timestamps,specifying prompts to generate at
particular seconds. For example, [0]a closeup pov view of a brunette woman in top of a man with his penis inside her throat, she has her head on top of the man penis with her mouth open [1]as he burst his penis inside her throat in fast up and down movements[4]the balls of the man bounces up and down, he holds her neck with his hand，she is looking at the viewer, povviewdeepthroat, headgrab, throatfuck, generating a specified prompt at the *th second. This way, a 5-second video could have 3 seconds dedicated to blowjob up and down. I'm just not entirely sure about the position of the timestamp within the prompt ([1] whether it should be placed further forward or further back in the prompt)😇
WAN version of this or something with the overspill effect would be great.
my ultimate deepthroat have this dataset inside
Hello there. With the kind of LoRAs you're publishing, I'm surprised you never released a "cum out of nose" version for wan. Or are angrydragon vids too hard to find? Keep'up 🙌
Details
Files
69Deepthr0at-throatpie-v1.1.safetensors
Mirrors
69Deepthr0at-throatpie-v1.1.safetensors
69Deepthr0at-throatpie-v1.1.safetensors
69Deepthr0at-throatpie-v1.1.safetensors
69Deepthr0at-throatpie-v1.1.safetensors
69Deepthr0at-throatpie-v1.1.safetensors
69Deepthr0at-throatpie-v1.1.safetensors
69 DeepThroat Blowjob ThroatPie v1.1 - Hunyuan Video - povviewdeepthroat,headgrab,throatfuck,noseonballs,gagreflex,penisswallow,headmove,penisrelease.safetensors
69 DeepThroat Blowjob ThroatPie v1.1 - Hunyuan Video - povviewdeepthroat,headgrab,throatfuck,noseonballs,gagreflex,penisswallow,headmove,penisrelease.safetensors