Workflow Venice AI Workflows and Models (DON'T POST REAL PEOPLE!!)

i keep getting censored on venice wan, do i need to pay for a subscription to get nsfw ? or is it nuked?
 
Your description of wan 3 confirms that it's limited the same way H3 is when it comes to nsfw. The models not being able to generate dicks, vaginas, and such on their own is a real pain especially for generating locally.
 
As I said, there's no problem generating a penis because I never had to show it an image, and there wasn't too much trouble (I'd say a 10% failure rate) getting the penis to penetrate the woman's vagina. And it does it well, like Grok.
To me, it's currently the best one for having someone undress and fucking them in the same scene.
+- 0.9$ for 20sec
 
Yes, passable motion and yes but not great if you want specifics
 
Wan 3.0 is unbelievably good. Gets incredibly expensive very quickly even with the 20% discount though.
 
Im having issues with "Wan 3.0 reference". most of my generations get stuck at 0% and then time out after 15mins, then it tells me I need to sign in again.
 
I saw "Lustify" underneath the image editor as a model but after I clicked on it, it disappeared. Anybody else see that model available?
 
I am so fucked...before I have tried a local version of facefusion...but holy shit...WAN and Seedance are just another level. Spend almost §400 the first weekend of use. I recommend to create a pdf with all rules that are important for different models like R2V or I2V. So I just feed grog with my pdf and it creates the story prompt using my general rules. Saved me a lot of time and the results are now much better than before. Also working on a prompt generator but this takes time.
 
How do I make the breasts physics seem real if I use wan 3.0 reference to undress the girl in original reference video? The bra makes her breasts appear pushed up
 
Derive secondary motion from established anatomy and the actual causal movement. Motion follows anatomy; anatomy does not
change to create motion. Use direction-aware inertia, gravity, acceleration/deceleration, damping, balance, posture, support and
contact. Different regions have different plausible responses; never apply one generic oscillation to the whole body.
CAUSE BEFORE EFFECT: acceleration or contact must precede the corresponding secondary response. After-motion loses energy
naturally and settles. Do not create endless oscillation, random jiggle or exaggerated elasticity. Movement Intensity controls the
primary action; secondary motion is then derived physically rather than independently maximized.
WARDROBE COUPLING: When clothing is present, account for its support, friction, stretch, folds and delayed motion. When absent,
remove clothing support without inventing exaggerated body behavior. Camera shake is camera motion and must never be
interpreted as body motion.

I would also reccomend to use a clear image of some breasts as reference
 
English isn't my native language; I was referring more to people who are assessed as minors because they look young, even though they are 23 or 24. Of course, I wasn't talking about actual children. You should have explained that better from the start :p And as I said, I highly doubt it, because that is the main source of problems for AI engines.
You don't mess with celebrities ;)
 
Kling 4.0 is coming with up to 30s single generation and 60s extended generation.

TBH, Kling 3.0 had the weakest NSFW performance in my testings (compared to SD, WAN and Grok). The model was censored and it wasn't really good at soft NSFW either.

But I have hopes about about 4.0 because they probably noticed the gooner market just like bytedance.

https://x.com/koltregaskes/status/2103019648321228916
 
Seedance 2.5 is working on 4K support.

Remind you, 1080p/30s video costs about 20 dollars on Venice. Imagine how much 4K is gonna cost.
 
You must log in or register to reply here.
Back
Top