I think it depends I don’t really see issues with it as tool I just don’t think ai content should be sold.
There are tons of issues with popular generative models outside of their use cases.
But I agree that AI content shouldn’t be sold (legally speaking, the seller doesn’t even own the rights to the product).
This true when talking about images but idk if it’s the same for code. You can make a program you own the program and it’s code. I don’t think funscript generators are trained off funscripts made by other people so would an algorithm and code made by something you own not be yours? (Spit balling here don’t take offense if it sounds dumb)
The training data doesn’t determine rights holder status of model users. It’s simply the fact that the product was not produced by a human that disqualifies it from legal protection.
From what I’ve seen that’s what a lot the arguments are about using other peoples ip to create something you claim you own.
It is true that they didn’t create the end product but they did create the code that made it without using any ip that they don’t own could it not be argued that the end product is still made by them?
Again, I’m talking about legal status and precedent.
Morally, I’d consider it quite different if someone trains a generative model purely off their own works. Extremely few people have actually done this tho.
3 Likes
Bravo!
Bravo!
Tbh, I don’t even mind if people want to use FunGen as a tool, the same way most OFS scripters use motion tracking when they feel the need to. If a script is good and someone’s generated it using FunGen, idgaf, I will buy it, but people just need to be more transparent, no one is going to bite your head off. A good script is a good script.
If a portion of a script was done by FunGen, whilst more attention was paid to complicated scenes, then share and talk about it in your notes, it’s impressive and shows you understand that AI just a small tool, but you’re still in control as the artist.
.
.
For example:
I would much rather someone like @Eternal_Enigma use FunGen (not saying they do!), if it meant they could spend more time on the fucking fantastic story, erotic animations, banging music, and most importantly, the entire point our haptic sex toy sub-culture: the erotic immersive experience.
Tangent on AI & scripting:
Once the automobile took over horse riding stopped being labour and became a sport, prestige, and an art form. the very machine that replaced the horse went on to give us Formula 1.
In ten years, when video AI can chew through the tedious scripting nobody actually enjoys, (better and faster than anyone) the tools that displace it will open the door to a Formula 1 of interactive content that’s grander, weirder, shit we can’t even picture yet.
Right now we don’t even know what we don’t know.
4 Likes
I think the general concern is that if a script is tagged AI-Assisted or the author discloses that fact it will get less sales overall, even if it is a good script. The counter to that, of course, is that there is hardly anyone making good AI-assisted scripts and disclosing it, so we have no way of knowing if the lower sales thing is true.
1 Like
Machine learning is just applying statistics and probability theory in general. Generative AI is a subset of that which focuses on specifically creating content (images, video, code, you name it). Effectively, machine learning is a means of generating content. They are not the same thing.
Realistically, mathematically, there is no other way to truly automate the process of action-based funscript creation to at least a satisfactory degree than by using machine learning. The possibilities of content are endless. I think people need to stop putting an equal sign between a means of application of mathematics and its result.
I can understand people wanting this niche to remain an “art form”, but as an engineer i look at the concept of a funscript and all I see is a rehash of the “human calculators” that cracked military encodings and got humanity to land on the moon. That was truly a feat of humanity 60+ years ago. Not a single person today does that anymore. I think the answer why is simple, really.
These funscript files only exist as a means to convey sexual movement in a video to a machine that moves a motor. They are literally a means to an end. Yes, you can make those devices dance on you in beautiful ways if you care enough. You can make them milk you dry, and people will love you for that. But that’s all this is. They are designed to move a device to make you cum. Maybe for some people that is equivalent to something an artist creates? I just don’t see it that way. The only way I can see it is a beautiful problem to solve with applied mathematics (machine learning). These exact same machine learning algorithms are applied to predict movement or detect things in all sorts of other videos and that’s just a normal thing these days. It’s not the same as an artist learning a literal craft and putting their mind and soul onto a canvas. It’s never going to be more than a specification for movement of a device…
But regardless, what people will feel is not something i want to dictate. I just see it as a means to an end, not the end itself.
1 Like
As a means to the end of making people cum machine learning is still significantly worse than the human calculators. I’m not sure whether or not I am hopeful about the possibility of perfect scripts at the touch of a button … my porn addiction is bad enough haha
For now maybe, although the quality seems to be advancing with every iteration of the underlying models/technologies. I’ve been going through my own stash and generating things and already i have scripts generated in seconds that are genuinely near perfect. It makes me cum just looking at them haha. Knowing i could finally enjoy even a part of my stash without being a human calculator for it is crazy. That’s all i’ve ever wanted since I found out about this. I don’t doubt that it will get progressively better over time. Just gotta improve these algorithms, the rest is mathematics.
1 Like
That’s optimistic! I have not had similar a experience with the tools I’ve used. If AI can generate workable/good scripts 10% of the time, that’s still something.
Maybe you could ask @k00gar for me since I pissed Roa off haha
Hahah, he already got tagged here plenty so if he wants to he probably will. I’m not directly interacting with k00gar anymore either, so it’s not up to me 
2 Likes
Machine learning is a means of predicting associations. Using it to generate content in a probabilistic rather than discriminative manner is generative AI.
They are indeed not the same thing. FunGen allows for the use of generative models to generate scripts based on input. It is therefore in the subset of generative AI.
Realistically, mathematically, you could create a discriminative rules-based algorithm to consistently output good funscript points based on visual input. It doesn’t need to use machine learning because you can if/then your way through the decisions of making a script based on a sequence of video frames.
I agree that we can automate funscripts, but I don’t think we need to claim any single methodology as the only correct solution. Or that we can allow people to be dishonest about the tools they use in the current climate where it is against the rules of the forum to not disclose a script being fully software generated.
The artistry of scripting is a fairly narrow discussion, and I do think generated scripts will eventually be good enough for most usage while there will remain a niche of creators whose style of scripting will be preferred by some people.
1 Like
Actually the rules-as-written as of right now are that you have to disclose any amount of AI use.
My interpretation of these rules are that if an AI tool like Fungen generates a single point of your funscript it should be marked as AI-Assisted
To be honest, i would love to see an if/then/else based approach to processing frames that could output the same quality on a 2D human porn vid as it could for a 2D furry porn animation or 2D pixelated hentai animation xD If someone can find a way to handle that infinite variety really well without using machine learning, my hat goes off to them, for real. I don’t believe for a second it’s a realistic approach though haha.
Of course people should be held accountable if they don’t disclose how they created a funscript that they share.
1 Like
Sure, but that rule will change any time. And MTFG not being disclosed is hardly sneered at.
1 Like
I hope the rule does change, it’s poorly written.
The variety is very much finite. A point is between 0-100 and movement can have three modes (positive, negative or neutral movement along the axis).
I have if/then/else’d audio based scripting, which definitely is child’s play compared to action based. Maybe if I had a budget I could try to organize an effort to solve that problem. Honestly, it’d probably make the most sense to approach a company like SLR if/once you get to the point of having a promising algorithm to build off of. But they might be too sunk cost fallacied into machine learning investment by now.
More broadly speaking, I also fail to see the underlying issue with using a probabilistic rather than a deterministic/discriminative approach to solving a complex real world problem. It’s also what our brains do unconsciously every second. By the same logic you used when saying the best approach is the one that yields the best results, you should not need to be biased towards preferring one over the other, so long as the best one wins.