I agree—this isn’t just a problem in the free section; even in the paid section, there are AI-generated scripts with no human editing whatsoever. I’ve been scammed once myself after paying for an AI-generated script. 
As for speed, I think some scripts are just too fast, resulting in an extremely poor user experience. Once the speed exceeds 450, it’s practically torture.
Yep. As shuny said above, there are even people who sell unrefined untagged AI generated scripts that don’t match the videos at all.
Great read, and I agree with most what’s been said. Personally, I don’t think anyone should be using AI trackers to generate scripts at all, except for personal use, and paid scripts should never be AI generated. If you’re charging for a script, you should have actually scripted it. @shuny404 getting burned on one is a good example of why.
One more thing that doesn’t come up much: these tools only seem to work on a pretty narrow type of video. I’m gay, and every tracker I’ve tried on videos with two guys has been useless. It locks onto the wrong body or spits out noise. From what I can tell, this seems to be because the models are trained on one specific kind of footage: attractive women and a man, shot from the usual angles. Anything outside that seems to fail. Gay content, amateur filmed shaky footage, other niche style stuff… I searched the site and gay + AI generated/assisted gave me exactly 5 results, which shows just how ignored gay people are in this space.
Sorry to be “that guy” again, but it does definitely get tiring having to be the one to bring this kind of stuff up, and working around tools that aren’t built with gay content in mind (especially when some of these tools charge an arm and a leg).
I also think PMV-style scripts get left out of conversations like this. Perfect sync to the action seems to be treated as the only measure of quality, since that’s just what ‘the majority’ prefers. But scripting to the beat, the music, or the mood can be just as immersive. I’d love to see more people get creative with it and try something different. Just another reason to do it by hand in the first place. 
Also, we should have a specific forum for asking for scripting help. The Help section exists, but it’s very general and mostly filled with questions about how to use devices/software. A dedicated spot for scripting questions would be a big help to newer scripters.
@VladTheImplier
I agree with this. There could be a “Beginner Scripter section” where people can post their first few scripts and get feedback and have “challenges” so if someone new does post they can be pointed in that direction. I think the best way to start learning to script would be to make a script of a video that is already done and compare but they will need the examples and that a “Beginner Scripter section” could house that.
A bit off topic but,
I believe this is purely a hardware limitation since I have a device that can actually hit much higher speeds naturally. That being said I do need to be better about making scripts that are more compatible with all devices across the board. I tend to really make scripts for myself and then release them as is afterwards, however It would be in my best interest to be more mindful that not every device can hit 800 or 450 like you said
I’m not gay and am currently devloping a tracking software for an SR6. I do want to take into consideration your complications. Is it just locking on to two dicks? or what is going on?
Realistically, I think as of now, there are going to be limitations given the variety of types of video/angles/camera quality… There are a ton of different factors when tracking a human body from a random video and translating that into accurate movement from a limited machine (thats not a knock on the SR6 at all. The thing is pretty awesome.)
But.. sorry I got lost on a tangent… I think, as of now, we’re at either scripting it completely by hand or having an AI give you a baseline to review and modify to personal preference. I am looking to improve that baseline automatically and the script creators can add their own flair if they want.
I am curious about OP’s post as well. What would you consider a higher quality script that I could test (preferably for free)?
Work smarter not harder but I get the people who love their craft and the effort it takes to put out their work. It is gratifying… Maybe people should look at tracking software as a “stencil” or like (lol) sometimes IKEA instructions..
It is there to guide and make the process easier.
I use beat generators and AI trackers. To lay out the foundation that I’ll be working on - later going through every single stroke point and adjusting it by hand. Sometimes tens of thousands of them. Tweaking, moving, adding vibrations, patterns, speed-ups and slow-downs. That is for stroker scripts. Vibration funscripts are built by hand from the ground up. Multi-axis funscripts I comission and collaborate with only established multi-axis scripters in the community (to whom I am massively grateful), knowing that they make it by hand.
The influx of non-human-tweaked scripts is noticeable. At this point when browsing the scripts categories - I intuitively pivot to my favorite creators, looking for their avatars and clicking their threads, knowing that I’ll find a good quality funscript in the post.
Maybe introducing some sort of “glow-up / stand-out” encouragement system is a solution:
You make high-quality, hand-made funscripts with a lot of detail?
You get a border around your avatar. Those can have levels too - with the amount of glow/shine/animation increasing with your reputation as a scripter. That way when people browse the catalogue and click on a post with a glowing avatar - they know they’re in for a quality product.
Also: adding a mandatory tag pick between “hand-adjusted-funscript” or “only-ai-made-funscript” in the required tags when publishing a post - also might help in solving the problem. Like right now we have to fill all the required tags for length, vr or non-vr, real or animation etc. A tag for funscript quality can help here.
Tutorials too - it would be helpful if there was a mega-tutorial thread, collecting all the possible resources for funscripting into one post, pinned at the top of let’s say - the general category. There are tons of great tutorials and guides available only if you search for keywords like “tutorial, how to script, funscript guide” and so on. It’s literally an under-a-minute search: you just have to look. Also - tons of established scripters share their funscripting insight just right there under posts for their own or other authors published works. With their knowledge just ending up lost to time. Maybe the administration could organize some sort of contest to encourage someone to collect this type of funscripting-tutorial database into one topic?
Just brainstorming here.
Sorry to bother you, but what’s that script in the last gif?
Should I create a new thread in the how to section that is for all things scripting?
This shall forever be known as…The Scripting Bible.
I will create separate categories for different actions - Bj, Hj, insertion,spitting, stroke lengths, and any others I can think up.
I didn’t read everything yet. But some say tutorials should be easier to find. I agree.
Right now, the howto section is a little messy. Maybe it could be separated like other categories are.
For example:
These seem the 4 categories that occur the most on that section. So a clean up there, may encourage people to check scripting tutorials out, because it’s easier to find.
A “scripting help” category / section could be nice as well. People can ask “how I script this?” “How can I do X better?” And so on. People can share their advice there. Many people use the general category to ask for scripting advice, which is not ideal I think. The Help section is good for asking, but I would create a dedicated section in the help category for just scripting advice. Cleaner cut. So questions like “how I reach trust level X?” don’t get mixed with scripting questions.
It’s already a rule that you need to disclose if you use AI generation at any point in your workflow for creating scripts with either ai-generated or ai-assisted. I’m hesitant to make it a ‘required tag’ though because that would mean creating a third tag for no ai used at all, which would mean… having to manually add it to all the older posts.
I noticed most of your posts don’t have these tags, which raises a bigger question: at what point does ai-assisted become basically a hand-crafted script? If everything gets adjusted by hand anyway, does the tag still tell people anything? And with no way to tell which scripts used AI, I don’t see how any of this gets moderated.
I think your proposal is great, so I’m voting for you.
Most of the problems just come from the fact that the AI lacks information. Not from the video, but in context how it needs to handle things, what intensity is desired, how the script does effectively result (note that a 10 length vs 20 length stroke is more than doubling in intensity because below certain lengths motion doesnt happen!). And it needs to know how a device will react.
Accuracy doesnt work, because its not how you would react.
Plain simple: the normal human touches to a script, are most of the time compensations for device limitations:
And there is the problem, most AIs are simply not well trained for porn. The only ones that have been tested a bit more is things like facial recognition, or just general events. Things that can map the general feel of a video and match these, or doing character replacements. Or they are made to generate the video. But barely any AIs are used to process scripts, and design a system to map a video to become a script.
And there are quite a few problems on this:
There are just too many flaws, and manual labour is almost always going to be most efficient for adding quality. AIs should only be used for things its known to be efficient: detecing timing (just having the up/downstroke pattern is fine if you can scale it to the proper length afterward), detecting camera motion and compensating this. Adding random noise afterward.
AI should be considered at most as an assisting plugin to help at a certain step. The more you ask of it, the more it will dissapoint you.
And a last thing: Test the result. If you notice its bad, dont release it. If you are consistent at dumping low quality scripts. I will just block you as user to filter out the trash. This means that even if you improve in the future, its already too late. Your bad reputation stays then.
Just saying AI generated is not a wildcard of being allowed to do this. Its only a bigger reason to quickly block you as user. And yes, 1 AI script you didnt flag as AI generated is to me an instant block.
My benchmark is simple: can i make something of similar quality in double the time that the video takes? If so, regardless of it being AI, i will deem it poor quality. And most AI scripts feel like that
I’m not using AI to track anything in the sense that it isn’t learning as it goes.. It is more of an algorithm than anything that tracks movements on screen.
getI hope to ge this software up and usable because it is worth it and would probably cut the labor time in half although it is processer dense.
In regard to AI restrictions for developing this software it is moreso how you frame it but I rarely run into guardrails. It is an engineering problem not a “get me off AI” problem. I’m an adult and not doing anything illegal with it.
I agree.. Don’t release something that hasn’t been tested both from a safety standpoint and.. like.. personal standards… Why waste someone’s time with a shitty product?
I do think I can produce a very accurate L0 script through tracking but, like you said, hand scripted will always favor a machine because.. we’re human and we know what it feels like from experience.
I’m working on a pretty decent workflow but it just takes time. I’m not sure which direction I wanna go with it.. Like have a built in script editor that is robust? or just have the export so one of the more established script editors can do the polishing for finished work.
No problem! It’s a 10 min script I’m working on (I think i’s about halfway done at this point). I feel like I’ve spent an eternity on this, but thats mostly because I’m bouncing between projects both free and paid. 
I like this idea! While some of us already have flairs on our profile pictures, they’re pretty small and don’t stand out much to the average user when scrolling the site. I think a border that shines (with varying levels) could not only attract attention to that person, allowing them to get more traffic on their posts, but could also incentivize people to make more hand-crafted (and better) scripts as they compete to get more interesting flairs. At this point, unless someone checks the leaderboard, or intentionally compares cheers, they won’t really get a feel for how experienced someone is unless they dig.
I like this as well. I generally put the “Hand-scripted” tag on my posts to give it a sort of legitimacy, but I wonder if it would be easy to just lie about the creation of your script and just post the hand-scripted tag over an AI post. Also, others can change or add a tag on posts in the community, do you think there might be some abuse of this ability where people mislabel others’ posts as “AI generated?”
I think this is a good idea, but I also agree with @Slibowitz that the howto - EroScripts section is pretty confusing to navigate for a first time (or even experienced) user.
This seems to be the best solution. Reiterating an idea I had earlier in the thread, I believe that combining this with a dedicated spot on the ES main/side menu would be the best way to ensure visibility and ease of use for all users.
Looks good, I’m looking forward to the full script being released.
I’m 100% with you on this one. I always reasoned that software can easily add fillers, but it’s very hard to remove without time consuming manual labor. Players often have a setting where they can automatically add filler strokes when the time gap without action is large enough.
