all
/
g
/
s
/
fur
/
art
/
ai
/
3dcg
/
f
/
rp
/
req
/
p2p
/
dis
/
lit
/
kaki
chat
GUROchan!
This list is WIP...
1
2
3
So use the classic menu
Link 4
Link 5
Link 6
Link 7
List 3
Link 8
Link 9
[
Settings
]
[
Home
] [
Catalog
] [
Search
] [
Thread list
] [
Stats
] [
Reports
] [
Watcher
] [
PMs
] [
Admin
]
Artificial Intelligence
[
Return
] [
Bottom
]
Posting mode: Reply
Name
E-mail
sage
Subject
Post
Comment
BBCode
B
I
U
S
Spoiler
Code
File
Animated GIF
Password
(for deletion)
Allowed file types are: gif, jpg, jpeg, png, webp, webm, mp4
Maximum file size allowed is 25000 KB.
Images greater than 255 * 255 pixels will be thumbnailed.
153
unique users in the last 30 minutes (including lurkers)
Switch form position
|
BBCode reference
|
Banned?
|
Quick reply
|
Post API
File:
fucking_example.mp4
(1.88 MB, 480x384)
Hide image
Hide image
▶
Degenerate LORAs
Anonymous
2026/03/22
(Sun)
14:51:33
No.
17302
[
Report
]
AS
▶
MODS: I know these aren't particularly on topic, but there just isn't a good place online to share these kinds of LORAs (that I know of).
What up degenerates, I trained a few gross ass video LORAs that some might enjoy. 3 are dog fucking-centric, and one is puking during blowjobs. The dog ones were trained on r34 cg, so they work *amazingly* well on video game characters and stuff, but they also do extremely well on real images 👀
NOTE: THESE ARE WAN 2.2 I2V LORAS, THAT IS ALL THEY WORK ON
That being said, would love to see what people make :P
https://mega.nz/file/P6wHRZTA#0fZgj5DgJp6dxtnfNCtPpe7htHgPbiOGRvoYol6uPdo
----------
doggy_lora:
Trigger word is doggy_lora. Trained on women getting mounted and fucked by dogs, mainly focuses on front/offside views and bent over/hands & knees/face down ass up positions. Mix and match to see what you like. Using the "blink doggystyle" lora as a helper can make it work much better with real images, put both at ~0.7 strength.
Prompt I got good results with:
Camera: The camera angle is a frontal view.
doggy_lora;
Beat 0: The woman bends over and gets on her hands and knees, and her clothes fall off, revealing her naked body and breasts.
Beat 1: ((As she lowers herself, a dog mounts her from behind)). The dog thrusts and humps against her ass roughly, causing her ass and breasts to jiggle from the impact.
Beat 2: She his crying and has a pained look on her face and her brows furrow as the dog fucks her ROUGHLY. The woman will vomit as she is fucked.
Beat 3: She will slowly lower her upper body until her chest is on the floor, all while the dog fucks her from behind. She will remain in that position until the end of the video.
----------
doggy_suck:
Trigger word is doggy_suck. Same as the previous, but focuses on POV dog cock sucking. AI struggles like crazy with the anatomy, so this one doesn't work as well as I'd like, but still very usable if youre fine with retrying.
Using the dgy_ck as a helper can make it give better results.
Decent prompt:
((The dgy_ck is attached to the dog's body.))
dgy_ck; doggy_suck; The same woman kneels down in front of a dog penis as her clothes fall off, revealing her naked body. She slowly slides the moist dgy_ck into her mouth and begins sucking it, her brows furrowing as it slides in and out of her mouth and moistens her lips.
The camera angle is a POV angle.
----------
dgy_ck:
Trigger word is dgy_ck. Simpler one just trained on dog dick to try to make the AI be less retarded about the anatomy. Works okay, I haven't tested it much outside of using it with doggy_suck.
----------
puke_lora:
Trigger word is puke_lora. Trained on a number of blowjob/face fuck videos where she pukes on the cock. Works fairly well, although its better if you either photoshop a cock in first (using a dick pic is fun af) or use the dr3aml4y helper lora to bring a dick in.
Decent Prompt:
A woman with a {scared | pained | aroused} expression is gagging on the penis as it begins face fucking her. she exhibits strong gag reflex responses, arching her back, shrugging her shoulders and frowning before vomiting a large amount of {thick | chunky | runny} {yellowish | brownish | whitish} bile and vomit, spraying out of her mouth and coating the penis. The chunky viscous puke coats her lips and chin with slimy vomit and drips down, (coating her breasts in chunky puke).
Viewing last 50 posts
>>
3
Anonymous
2026/03/23
(Mon)
19:38:01
No.
17331
[
Report
]
▶
Good stuff, hard to find stuff like this. Also what is your process on how to train your loras? Also a request, can you do horses next?
>>
4
Anonymous
2026/03/24
(Tue)
10:36:21
No.
17360
[
Report
]
▶
>>17331
>>
5
Anonymous
2026/03/24
(Tue)
15:38:01
No.
17370
[
Report
]
▶
>>17331
The process isn't exactly difficult, its just tedious. I did it with ai-toolkit & renting GPUs on runpod cause it takes forever even on a good GPU.
Basically like:
- Find ~10 videos of the thing you want to make (3-6s long). They have to be similar enough that it won't just confuse the training.
- For each video, make an accompanying text file with the same name that describes it and has a trigger word if you want one (so vid_1.mp4 + vid_1.txt)
- Crop each video to focus on the theme of the LORA (easy in clipchamp)
- Optional: if you want the "blink" type loras to use with I2V, add a ~0.3s image of the SAME person/character to the beginning of the video (this is the hard part for a lot of training videos, you need an image of them *not* doing the thing in the video)
- Upload all the videos & text files to a dataset in ai-toolkit
- Configure a training job (there are tons of settings that are tweaked for GPU, lora strength, etc. Wont get into that)
- Train the lora (For a ~2000 step training, this takes about 2.5 hours on the super high end runpod GPUs)
As for doing horses, I *could* try that, but I think there might be too much motion/extraneous stuff for it to work well. I might try like horse cock BJ first since thats a bit simpler motion wise. Plus other existing BJ LORAs can help push it in the right direction
>>
6
Anonymous
2026/03/24
(Tue)
21:43:17
No.
17381
[
Report
]
▶
>>17370
Thanks for the reply, seems simple enough. I always wondered how they do the "blink" style loras, so thanks for that. Is a 5090 good enough to train on wan and do you have any links to a good config for training settings?
>>
7
Anonymous
2026/03/24
(Tue)
22:15:25
No.
17382
[
Report
]
▶
Really been trying to find some degen loras for chroma. Any leads?
>>
8
Anonymous
2026/03/25
(Wed)
00:22:50
No.
17386
[
Report
]
▶
>>17382
Never understood the appeal of chroma, it feels worse or on par with illustrious, but with less lora support. Anima looks promising though.
>>
9
Anonymous
2026/03/25
(Wed)
15:40:26
No.
17416
[
Report
]
▶
>>17381
No problem, it definitely seems like people try to gatekeep how stuff is made for sure.
You *can* use a 5090 technically (thats what I have), but it will be insanely slow. Like...8-12 hours of full tilt processing probably.
I put a config I used here:
https://pastebin.com/q6yDrJ9x
Usually I just swapped train_high_noise and train_low_noise to do one at a time because its easier on the VRAM. Just note, its a lot of trial and error to see what works and what doesn't.
>>17382
Unfortunately not, I haven't really used chroma at all. You can check civitaiarchive though, it backs up deleted LORAs so you might find some that got removed.
>>
10
Anonymous
2026/03/27
(Fri)
04:39:38
No.
17450
[
Report
]
▶
>>17302
absolutely cinema
>>
11
Anonymous
2026/03/27
(Fri)
12:52:23
No.
17462
[
Report
]
▶
>>17331
Horse fuck is now training. We'll see how it goes o7
>>
12
Anonymous
2026/03/27
(Fri)
20:54:00
No.
17480
[
Report
]
▶
File:
horse_fuck.mp4
(1.31 MB, 480x384)
Hide image
Hide image
▶
It honestly worked better than it had any right to, lmao (also with bonus belly bulge).
https://mega.nz/file/zvYGiQgA#x8CqIwW2fF--8KYzsvTc68MBPAvDffcPs0NAYcxTOtg
It works decently on IRL images (~0.85 strength seems best), but not as good as the other loras unfortunately. Fuck around with it, see what you can get. I still haven't found great settings.
>>
13
Anonymous
2026/03/27
(Fri)
20:54:48
No.
17481
[
Report
]
▶
>>17480
Also these cost like $30-$40 to make, but I do it for the love of the game 😭
>>
14
Anonymous
2026/03/27
(Fri)
22:28:01
No.
17483
[
Report
]
▶
>>17481
Thank you for your service 🫡
I'd train on my 5090, but 8-12 hours seems really steep investment for something that might not turn out good.
>>
15
Anonymous
2026/03/27
(Fri)
22:36:59
No.
17484
[
Report
]
▶
>>17483
Yeah, it takes way too long. And that would be for like high or low noise, youd still have to do the other.
You can have it save a lora every X steps for testing, but it only gives you a vague idea of how well its going.
>>
16
Anonymous
2026/03/27
(Fri)
23:04:07
No.
17485
[
Report
]
▶
I'd love to make some actual guro ones, but there's just no good videos I can find to train it on.
>>
17
Anonymous
2026/03/28
(Sat)
14:52:08
No.
17527
[
Report
]
▶
>>17485
What if we generate gore/guro content using Imagine ( ie. Grok) then train a lora on that? It has been obviously trained on horror/splatter movies and alike and it’s the closest you can get without using actual real dataset
>>
18
Anonymous
2026/03/28
(Sat)
14:53:47
No.
17528
[
Report
]
▶
>>17527
Or training directly on movies/ etc.
>>
19
Anonymous
2026/03/28
(Sat)
15:37:39
No.
17531
[
Report
]
▶
File:
horse_prone.mp4
(1.59 MB, 544x432)
Hide image
Hide image
▶
>>17527
True, movies might work but you're limited on the actual acts you can get and (if this post is any indication) I'd prefer degenerate shit.
Unfortunately I'm pretty sure training on AI just reinforces the AI feel; kinda like when you run the same picture through ChatGPT a dozen times or whatever
Also, anothaOne.gif
https://mega.nz/file/27w0mZrJ#BpA5GokLU79u9MvvnxsJXxPnKFHQwIMgTbucWbL3Nnk
uses horse_prone as a trigger word, the last one uses horse_fuck, forgot to mention.
>>
20
Anonymous
2026/03/28
(Sat)
15:39:02
No.
17532
[
Report
]
▶
>>17531
Also to be clear, all the examples are the literal first generation it did. Didn't hunt for good ones, just generate -> save, lol
>>
21
Anonymous
2026/03/28
(Sat)
17:16:45
No.
17533
[
Report
]
▶
>>17531
If you're into degenerate stuff, may I request insect/oviposition/egg laying next?
>>
23
Anonymous
2026/03/29
(Sun)
18:40:57
No.
17550
[
Report
]
▶
>>17533
Hmm, not a bad idea. I'll have to look through some and see if there are any examples that seem appealing
>>
24
Anonymous
2026/03/31
(Tue)
15:21:42
No.
17620
[
Report
]
▶
Great work OP, if we can help in any way let us know. You are basically the only one doing both something useful and not of garbage tier quality in this genre
>>
25
Anonymous
2026/03/31
(Tue)
15:23:26
No.
17621
[
Report
]
▶
>>17483
If I had a 5090 it would be running almost 24/7 to train loras
>>
26
please do more of these beast lora
Anonymous
2026/04/01
(Wed)
00:56:10
No.
17659
[
Report
]
▶
I love you
>>
27
Anonymous
2026/04/02
(Thu)
00:15:23
No.
17684
[
Report
]
▶
>>17621
Depending on the type of lora/content I might be willing.
I'll say now, no to scat content.
If you make the dataset and provide a zip I'll try training it.
>>
28
Anonymous
2026/04/02
(Thu)
15:31:27
No.
17698
[
Report
]
▶
>>17659
And I love you, random citizen
>>17620
Honestly just finding the reference videos is the longest and most difficult part, if anyone wants to find good sets of videos then I don't mind making more.
For reference, look at this guys work on r34:
https://rule34.xxx/index.php?page=post&s=list&tags=thebartender+video&pid=0
These are a lot of the ones I used on the beast loras. Notice how so many of the videos have the same position/camera angle? That's the kind of stuff that works well (that I've found at least).
So you basically need "thing you want to see + numerous videos of it that are similar enough to train well"
>>17533
I looked a little, but ran into the issue above T.T
Just cant find many videos of the same type of thing to make it work
>>
29
Anonymous
2026/04/02
(Thu)
21:45:30
No.
17728
[
Report
]
▶
>>17698
Unfortunate, half expected so since its niche. I have an idea for maybe hot iron branding? Im sure branding videos exist for animals and other inanimate objects, would a general lora for "Branding ____" be possible?
>>
30
Anonymous
2026/04/03
(Fri)
17:40:30
No.
17769
[
Report
]
▶
>>17728
Hmmm, maybe. Usually I check r34 because it's easy to find a bunch at once, but the closer you get to guro stuff the less there is there. I'd be concerned with fur and stuff messing with the training, but I'm not actually sure how much it would care.
>>
31
Anonymous
2026/04/05
(Sun)
07:35:02
No.
17816
[
Report
]
▶
>>17684
Would love to see a couple of things from this if you don't mind requests.
Would definitely like to see Ashley puking on dog dick! And perhaps some gore of her? Maybe getting her guts spilled out, however you would like.
>>
32
Anonymous
2026/04/06
(Mon)
17:23:07
No.
17867
[
Report
]
▶
File:
vid_00034.mp4
(2.07 MB, 544x432)
Hide image
Hide image
▶
>>17816
Sure thing.
Unfortunately I can't really do guro ones, the models don't know how and there aren't any good LORAs for it (that I know of)
>>
33
Anonymous
2026/04/06
(Mon)
17:23:24
No.
17868
[
Report
]
▶
File:
vid_00001.mp4
(1.93 MB, 544x304)
Hide image
Hide image
▶
Bonus remake Ashley
>>
34
Anonymous
2026/04/07
(Tue)
06:10:49
No.
17886
[
Report
]
▶
>>17867
Hot stuff! Thank you. Looking forward to see more of Ashley getting dogged or horsed from this thread. Unfortunately my PC isn't nearly strong enough to do any sort of AI generating myself. Though I hope you don't mind I make another request?
>>
35
galathien
◆0AdiozdanE
[
PM
]
2026/04/07
(Tue)
07:18:34
No.
17888
[
Report
]
▶
>>17867
It may not be guro, but I like it.
>>
36
Buggy
2026/04/08
(Wed)
12:31:48
No.
17935
[
Report
]
▶
>>17302
We really need a bestiality board.
>>
37
galathien
◆0AdiozdanE
[
PM
]
2026/04/08
(Wed)
19:47:25
No.
17943
[
Report
]
▶
File:
horse-thumbnail.png
(1.12 MB, 928x928)
ImgOps
Hide image
ImgOps
Hide image
▶
As long as we're posting some AI beasty stuff...
https://files.catbox.moe/qia13w.mp4
>>
38
Anonymous
2026/04/09
(Thu)
05:30:36
No.
17952
[
Report
]
▶
>>17943
You're dancing with the devil
>>
39
Anonymous
2026/04/09
(Thu)
15:21:57
No.
17967
[
Report
]
▶
>>17943
the link seems to be down
>>
40
galathien
◆0AdiozdanE
[
PM
]
2026/04/10
(Fri)
04:24:26
No.
18000
[
Report
]
▶
>>17967
> the link seems to be down
It's catbox.moe link. I use it for the anonymity and the easy of use, for both me and anyone clicking on the link.
catbox.moe gives you immediate access to files, no CAPTCHA's, no questions, no attempts to get you to subscribe to a file sharing service, etc.
What catbox.moe lacks, however, is 24/7 reliability. And the admins there are free, of course, to delete files they find and don't like.
The link is working for me now, however, and will probably work later for you.
>>
42
Anonymous
2026/04/10
(Fri)
12:21:12
No.
18011
[
Report
]
▶
>>17967
If you use a VPN, change the c in catbox to an f, and it should work so long as the file exists
>>
43
Anonymous
2026/04/10
(Fri)
16:21:26
No.
18023
[
Report
]
▶
File:
vid_00001 (1).mp4
(1.45 MB, 544x304)
Hide image
Hide image
▶
>>17886
Enjoy!
>>
44
Anonymous
2026/04/12
(Sun)
06:14:50
No.
18088
[
Report
]
▶
>>18000
checkpoint?
>>
45
Anonymous
2026/04/12
(Sun)
06:40:59
No.
18089
[
Report
]
▶
>>18023
Niiiice. Could we get a video of her with massive "ballistics" to titfuck a dog or horse?
>>
46
Anonymous
2026/04/12
(Sun)
11:32:30
No.
18093
[
Report
]
▶
>>18023
Yeah! And let horse jizz in her. And her piss after horse shaft squeezed her bladder.
>>
47
Anonymous
2026/04/13
(Mon)
09:21:24
No.
18132
[
Report
]
▶
File:
ComfyUI_00009_.mp4
(678 KB, 640x640)
Hide image
Hide image
▶
Thank you very much OP i had been banging my head for weeks trying to get decent videos to generate, mostly just came out as static or ungodly abominations, Will continue to experiment.
>>
48
Anonymous
2026/04/13
(Mon)
17:12:45
No.
18136
[
Report
]
▶
>>18089
>>18093
Lol, I love the enthusiasm, but the AI doesn't know how to do that. The stuff I linked "teaches" it how to do those specific things, but it can only do those unfortunately.
It does work great for basically any CG or video game character though.
>>18132
Nice! I'm glad it helped, its been a lot of fun experimenting. All of mine are usually just euler_ancestral sampler with simple scheduling. With a lightx2v lora it only needed ~7 steps, 4 low and 3 high.
>>
49
Anonymous
2026/04/19
(Sun)
11:51:23
No.
18305
[
Report
]
▶
>>17480
can I ask for prompt for this lora?
>>
50
Anonymous
2026/04/20
(Mon)
15:38:20
No.
18351
[
Report
]
▶
>>18305
Sure thing, I usually use something like this:
horse_fuck;
The view is from the side. Her entire body and face will stay in frame at all times.
The same woman smoothly removes her clothes and is completely nude, exposing her small breasts while bending over, facing the {left | right}. She will look at the viewer at all times.
As she bends over a large animal mounts her from the back, while staying partially offscreen, its back legs positioned behind her ass while its front legs are above her head. The animal begins fucking her asshole roughly. The animal's huge, thick penis is thrusting inside of the woman repeatedly. The animal's cock pushes completely inside her, then comes back out. Her abdomen bulges greatly with every thrust and she is pushed forward slightly. Her naked breasts jiggle and bounce with every hard thrust.
>>
51
Anonymous
2026/04/20
(Mon)
23:05:01
No.
18357
[
Report
]
▶
Im not sure if this is OP, but theres another horse lora on Civitai for wan i2v. Make sure to scoop it up while you guys can. It its not OP, maybe he could get into contact and collaborate?
https://civitai.red/models/2560303/wan-22-horse-lora-i2vt2v
>>
52
Anonymous
2026/04/21
(Tue)
03:00:46
No.
18362
[
Report
]
▶
>>18357
That is not me, but thanks for the heads up! I'm interested in seeing how well the two will stack
>>
53
Anonymous
2026/08/22
(Sat)
23:26:08
No.
25246
[
Report
]
▶
>>17943
>>17943
Is there more of this? Link or telegaurd?
>>
54
galathien
◆0AdiozdanE
[
PM
]
2026/08/23
(Sun)
03:08:07
No.
25263
[
Report
]
▶
I wonder what the future of this thread will be for MiniMax H3?
MiniMax H3 is pretty capable as-is, especially compared to basic Wan 2.2, but there are some things MiniMax doesn't do well that I can definitely see being helped with a few LoRAs.
The problem is that MiniMax (MiniMax Group, the company) is aggressively seeking takedown notices for NSFW LoRAs designed for MiniMax.
I've never done much LoRA training myself except some not-so-successful attempts at character LoRAs. But I might have to learn how, and make some LoRAs myself at this point. For some sexual content, training material is abundantly available. For gore? I have no idea what one even uses for input to make a good gore LoRA. (Short of becoming a serial killer oneself to create one's own very graphic inputs.)
Has anyone in this group found any good LoRAs for MiniMax yet?
[
Top
]
Delete post: [
File only
]
Password:
Style: