Shameless plug for my deviant art page https://www.deviantart.com/le-razzz
Follow me there for great Ai video content.
V4 is now out. New concepts
no bra pokies
anal fisting
vaginal fisting
big hanging boobs
bondage
spread pussy
futanari
anal prolapse
I recommend .7 for the Lora and .3 for unconditional. Euler look more realistic now so you can get faster realistic generation.
V3 add a bunch of sexual position concepts
Here's the full list of all the concepts so far:nude woman
nude man
pov vaginal sex
pov doggy style vaginal sex
reverse cowgirl vaginal sex
sideview doggy style vaginal sex
doggy style anal sex
pov anal sex
reverse cowgirl anal sex
pov blowjob
blowjob
sloppy blowjob
handjob
titjob
vaginal dildo
lesbian pussy licking
huge natural breasts
face cum
hairy pussy
pussy insertion (name your object)
puffy nipples
big fake boobs
upskirt
ass spread
creampie
gaping pussy
gaping anus
self fisting
Note: There might be a bit of pussy and anus confusion sometime. In that case put a bounding box where the other hole should be.
V2 add men and a bunch of new concepts.
Don't go higher than .90 or it will start creating bad quality image.
Keep the steps at 18 instead of 12 as you may get generations error from time to time.
Sweet spot is between 0.50-0.90.
res_2s = best definition
res_2m = most realistic and faster than 2s
euler = fastest but less realistic
https://huggingface.co/RazzzHF/Realism_Engine_Ideogram_4/resolve/main/rei4_v2_workflow.json
V1 - First RE Ideogram 4 release. Trained on female DS only so far. Male is coming in the next versions.
Sweet spot is between 0.50-0.60. Higher will degrade the quality.
MAKE SURE YOU USE JSON PROMPTS.
res_2s res_2m is recommended.
Description
Added sexual poses.
FAQ
Comments (31)
simple solo female fingering is always overlooked in all nsfw models. wonder why.
I have a decent-ish dataset of traditionally captioned poses if you're interested in adding. Not sure how the captioning should differ for idogram
The fun thing with ideogram 4 and a multi concept Lora is that since it knows what is pussy, pussy insertion and fingers, you don't even have to train for it. https://civitai.red/posts/29163620
@razzz almost. I'm looking for some full spiderman web slinging pose /''\` type penetration. but this is a start
V3 is godlike, you're the GOAT Razzz
V3 IS AMAZING!!
any special numbers to put these at or still .8ish?
Yes, you can play with the values a bit. Bringing down the Lora value will improve the aesthetic but make harder to do the more complex concept.
For the unconditional one higher you go the cleaner is the image but you lose a bit in realism.
For the next version you should teach it how to do spread pussy, including all of the little bits and stuff so its perfect.
I haven't actually tested v3 yet, but I did just download it so I will play around with it in a bit. But since spread pussy wasn't on the list of concepts I assume it wasn't included.
Godlike
No big fake boobs preview for v3?☹️
No but don't worry, everything from V2 is available in V3.
Well... It is kinda difficult for me to use v3. The same prompt gives different result, looking badly :/
OP is running the Lora on each model, normally at lower strength on the unconditional model. Also, they're using the Ideogram Scheduler. I personally do not put any loras on the unconditional model due to memory constraints (but also find it doesn't significantly improve quality with it on anyway). I use it at 0.4-0.6 strength depending on how NSFW the photo is. I'm personally seeing better results each iteration (V3 > V2 > V1).
I will say V3 seems to lower the variety on some topics -- might be that some parts of the training set have hit their limit. However I have had some good success reintroducing variety by simply using the V2 lora along with V3 without any loss of fidelity. In particular all my "mature boob" like photos lost some variety in V3, and blowjob tests became more stiff to anatomy (i.e the bounding boxes have to be correct or it'll be more likely to fail) -- most of the other tests I ran kept the same variety as before. I'd say it does seem like the V3 version has hit it's max on a few of the topics.
Please do the concepts (anal full nelson, interracial gangbang and cheek blowjob) for a future update, that would be amazing!
I'm pretty sure anal full nelson and interracial gangbang can be done with V3 already.
@razzz Yes, I figured... But I'm referring to more targeted/specific training, with a more in-depth dataset on these concepts I mentioned, to generate even more impressive results. :)
bro is upgrading the lora every day lol great work
nice
Damn. I have very high standards when it comes to nsfw ai, but it's undeniable this is no longer a gimmick
v3 became unstable. It stopped following attention blocks. LoRA weight 0.4 - 0.6. CFG 3.5 - 7.
try 0.6 cond - 0.5 uncond and cfg 5
Is it just me, or did the women get a lot less attractive in v3? Nope, I'm an idiot, this lora is amazing. That 1NNIE pussy one is awful though. Makes them into frump-town.
The innie one just seems to have a small dataset. It actually works, just doesn't play nice with others and has low variety.
The base model doesn't understand common sex toys, could you maybe cram that into the model by chance?
Fantastic lora. Can you please add some swimwear tanline training data or a separate lora? That would be great
is it possible to have some 2 v 1 ? like spit roast?
thanks for your hard work btw
Great work! Can you please teach it different body and breasts sizes? Now it biased towards an average build with medium boobs.
Amazing work! Will V4 be coming at some point soon and what changes/improvements do you think it will bring?
I must admit this is a great project. In the past, any image generated by any model would be detected by large language model analysis, but this time it successfully fooled gemma-4-31b and qwen3.6 35b.👍
This lora affects the face, especially Asian women.
Looks like we don't have an active mirror for this file right now.
CivArchive is a community-maintained index — we catalog mirrors that volunteers upload to HuggingFace, torrents, and other public hosts. Looks like no one has uploaded a copy of this file yet.
Some files do get recovered over time through contributions. If you're looking for this one, feel free to ask in Discord, or help preserve it if you have a copy.