I built an AI edging system that reads my body and controls the Keon

DIY
21

Great questions! Happy to share some technical details.

1. Nogasm comparison: Honestly, I can’t give a fair comparison. I built a Nogasm years ago but my session was a failure due to impatience - never gave it a proper chance. The technology itself is still very interesting to me. I’m actually thinking about how to integrate that type of sensing without the plug, because the principle keeps bouncing around in my head. If you have ideas on that, I’m all ears.

2. On-board AI + ML review: Yes, everything runs on-device on the Body ESP (SC01 Plus). The session logs are CSV files that include an Event column tracking states like NORMAL, PAUSE, WARMUP, ORGASM, COOLDOWN, and AI_CONTROL. These feed directly into the ML training pipeline. I have a 3-slot model system (SLOW, FAST, EXPERIMENT) and the AI builds “experience” from logged sessions - 100 samples = 100% experience. It also logs when I override the AI with the nunchuk, so it learns from my corrections.

3. PONR tracking - threshold vs modulation: It’s actually both. The system uses gradual intensity modulation based on stress level: full intensity at low arousal, reduced to 85% at medium, 70% when getting close, and complete stroke skipping when critical. But there’s also a hard threshold response: when edge is detected, the AI drops aggressively to a low level (not gradual - one step back = ruined orgasm, so it backs off hard). There’s also a PONR timer: if sensors keep rising after a cooldown drop, the AI will trigger orgasm mode to prevent a ruined one.

Your point about estim signal interference is exactly why I went the non-pelvic route. Keeping the biometrics separate from the stimulation devices opens up a lot of possibilities.

Really glad to hear you share the vision on non-commercial and attribution.

1 Like
22

The Nogasm plug principle is based on an inflated plug with connected tubing. It turns out you can just use the tubing:

Prause, Nicole & Arnold, Rudolf. (2019). Anal pneumatics: Tool for the verification of orgasm contractions across genders. Submitted for Publication.

That paper created a simple plug design where the tubing spans the anal sphincter. In my system, I used the same principle to surround the ‘neck’ of an anal estim toy with tubing, allowing accurate pressure sensing with a connected SparkFun MicroPressure Sensor.

I’ve had some success with modifications to more traditional toys - anything with those small insertable ‘lipstick’ size vibrators. In place of the vibrator, you can use a small printed tube to create a cavity that connects to the external tubing. The material of the toy means its soft and flexible like the tubing. Hard to describe but I can share some photos in a few days.

The basic idea for both is the same - you need a small, compressible volume spanning the sphincter connected to a pressure sensor. No inflation or pumps required. This means you can use a small plug if desired (paper above has a small plug designed to be used with condoms during patient trials). I’ve been dwelling on the plug-free sensing as well and have yet to come up with anything.

Thanks for the info on your AI/ML, I’ve ordered some parts to try it out out. I had been integrating heart rate with a smart watch but hadn’t had much luck, I suspect combining it with GSL as you have will give much better data.

Your PONR tracking is similar to my implementation, that’s reassuring I was on the right track! I have a hard cutoff to prevent orgasm if the pressure spikes too high. For the modulation I paired pressure with intensity using an inverse quadratic relationship - quite slow reductions as pressure starts to build from from low to moderate, with faster reductions in intensity as you move between moderate to high/edge. Fast ramp times for this were mind-bending combined with estim - felt like a constant wave of being brought right to the edge and back down.

I’m very curious how arousal modelling looks compared to pelvic floor pressure. In my experience, estim let me ‘force’ pleasurable sensations as part of the system. So even if arousal wasn’t very high, the system could still force edges.

“Your point about estim signal interference is exactly why I went the non-pelvic route. Keeping the biometrics separate from the stimulation devices opens up a lot of possibilities.”

My modified electrode only uses the internal surface (prostate area) connected to other electrodes - so its not capable of creating a sphincter contraction just with estim. Those start to happen as you approach PONR. What I found fascinating was at high intensities, it felt like the prostate stim WAS causing full contractions, but the pressure sensor stayed low. There were other muscle groups involved being stimulated that was pleasurable, but not actually causing the type of contractions I expected.

1 Like
23

Very interesting info about the simplified tubing approach! I didn’t know you could skip the inflation part. That research paper and your modifications make a lot of sense.

To answer your question about arousal modelling vs pelvic floor pressure - honestly, I don’t know yet which one is better. Pressure sensing seems like a very reliable method for sure. But I’m thinking: maybe a combination could be even better? A less precise pressure measurement combined with HR, GSR, temp, and breathing might give more complete data than either approach alone. Something to explore.

Your idea about printable designs with tubing is interesting. I actually have experience with TPE printing, so something with an integrated tube could work. But that’s for version 2.0 - not there yet.

About the estim side of things - I have zero experience with that. I’m a bit old-fashioned and prefer the classic up-and-down motion like the Keon provides. So I can’t really compare how arousal modelling would work with estim specifically.

Would love to see those photos when you have them!

1 Like
24

I should clarify - the paper and my method use 3D printed probes or parts of probes - the tubing is just standard plastic tubing wrapped round. TPE could work but I’m not convinced it would be body safe. The paper uses condoms so no concerns there, and the 3D printed part of my estim probe is a small sheath for the neck of the toy that sits under the tubing - so no contact either.

I’m expecting GSR and HR/Resp parts to arrive later in the month so I can see how it compares. How effective have you found your system at edging so far?

I’ll get some photos later of my approach for estim and the ‘lipstick’ style sensor for other toys.

1 Like
25

Detail on my pressure cavity attachment for pressure sensing using generic toys designed for insertable vibrators. I noticed there were a lot of toys that used this same shape and realised the flexible necks could act as a pressure cavity themselves.

Pressure Cavity Attachment
Pressure Cavity Attachment3952×2494 1.81 MB

In my experience these worked as well as my tubing setup (photos to follow). However, you may need to tweak the design to get a snug fit for your toy. This could be an accessible way to access pelvic floor data without needing the full inflatable plug approach used in the NoGasm and similar devices.

STL will be uploaded tomorrow for those interested.

1 Like
26

You’re right about TPE - probably not body safe for this. Good point about the condom approach from the paper, that’s a smart workaround.

Your pressure cavity attachment is a really clever and well thought-out design. I like how you identified that the flexible necks of those toys could work as a natural pressure chamber. Makes it much more accessible than the full inflatable plug setup. Looking forward to seeing the STL!

As for how effective my system is for edging so far - honestly, I’m still in the testing and tuning phase. The sensor data looks promising and the AI logic is working, but I need more real sessions to really dial it in. The ML needs training data to learn my patterns. Ask me again in a few months! :grinning_face_with_smiling_eyes:

Currently I’m busy designing the Pip-Boy style armband enclosure. Taking a while though - 3D design isn’t really my strong point.

Curious to see your photos when you have them. Always interesting to see different approaches to the same problem.

27

As a user yes I would be very much interested in this project. However I do not have the technical knowledge of the ESP32 to contribute. I can assemble things and stuff like that but the programming side, its beyond me. This is honestly something Ive been wanting, Im kinda turned on by the idea of turning it over to an AI that I cannot predict. I would love to see how long I can truely go without stopping.

2 Likes
28

Just wanted to give a small update. I’m letting all the positive reactions sink in for a bit - didn’t expect this much interest to be honest, seeing all the likes coming in!

If things go quiet from my side, don’t take that as a sign I’ve abandoned the project. It’s actually the opposite - it means I’m busy working on it. Free time is scarce for me, so progress happens when it happens.

I’m still figuring out what direction to take with this. Whether to keep it personal, share the concepts, maybe let a few people test it… haven’t decided yet. But I appreciate all the enthusiasm and ideas. It’s motivating to see others are interested in this kind of thing.

I’ll keep you posted when there’s something to share.

2 Likes
29

This is a refreshing attitude. Heck, I could even call it noble, but let me offer an alternative viewpoint. It seems like your goal, other than enjoying a fun side project, is to give something of use to the community. Well… the best way to do that is to sell us something. Not everyone will have the ability or will to assemble one of these even if you provide the code, parts list, and instructions. The way I see it, there are a few ways you can approach this:

  1. Completely user-serviced DIY: This is what you’re already imagining. The user is required to source all parts, assemble, solder, program, and troubleshoot everything. This is the cheapest and most tinker-friendly option, but also the least accessible.
  2. Build Kit: You source the parts to make sure everything necessary is included, maybe preprogram anything you can, mark the kit up €10-30 or whatever would make doing this worth your time, then ship it off as a fun little project for the end user. Adafruit Industries among others does things like this. I’ve assembled something much simpler once and building your project might be worth doing, but I’m a man of modest electronic skill and troubleshooting if everything doesn’t work perfectly the first time might be beyond me.
  3. Selling a Finished Product: Exactly what it sounds like. This would be the most expensive option for the end user, but also the most accessible. Of course, if you go this route, people would expect to be provided some form of product support, which it seems you aren’t interested in. Still potentially worth considering.

Also worth considering is that you can sell stuff and still leave the schematics and code open source and available for anyone who does want to build/improve a unit on their own, so a combination of 1 and 2 or 3. Basically, buying the expensive thing would just be an option rather than the only way to experience what you’ve created. Isn’t that what the OSR guy is already doing? Regardless, making money on something you’ve worked hard on isn’t evil, you can certainly do it ethically.

It seems you’re already realizing this, but I’m still going to say it: you’re underestimating the utility of your idea. You think you’d sell 10 units at €100 each? I’m in the US and if you made this t-code compatible and functional out of the box I’d gladly pay you $150/€128 right this second before shipping. Put some of that Apple craftsmanship on the looks and interface and $200/€171 isn’t unreasonable, though you can of course set a lower price as it suits you. 10 units sold is an estimate so humble it almost feels like part of a comedy routine.

No matter what you do I’ll keep monitoring this thread, though. What a neat idea.

2 Likes
30

Wow—this sounds awesome! Over the last week or so, I started researching whether something like this was even possible. It turns out it is, but without much technical background I was feeling a bit lost, and AI was only so helpful.

My plan has been to build a Nogasm/EOM3K as my first hands-on project to get into tinkering. My second project was going to be using Python to create an edging app that listens for keywords (using Whisper) to interrupt a toy running a funscript, pause for a random cooldown period, and then taunt using a script before restarting stimulation.

My hope is that, eventually, I can combine these ideas with a live AI to take full control of a session.

2 Likes
31

Cool to hear you’re diving into this! Starting with a Nogasm/EOM3K is a solid first project to get your hands dirty with the basics. And the Python idea with Whisper for voice-triggered interrupts sounds interesting - that’s actually not too far from what I’m doing with voice control.

Keep at it. The technical stuff can feel overwhelming at first, but it clicks eventually. AI can be a great helper, but yeah, they can also be very stubborn and annoying sometimes :grinning_face_with_smiling_eyes: Just keep pushing until you get what you need.

I’m pretty busy working on my project right now. Also looking into whether Patreon might be something - still figuring that out. We’ll see where it goes.

Good luck with your builds!

1 Like
32

+1 to everyone else’s responses, this looks great!

My only recommendation is no matter what path you choose, the best thing to release publicly would be any biometric correlation data you develop. Being able to correlate the readings from the sensors you’ve selected to orgasm would be a great contribution to the larger community!

2 Likes
33

Looks really, really cool. And it’s amazing you got the ml to run on an esp32. If I had a suggestion, I’d suggest that the community could contribute tagged data, so the ml model gets better

1 Like
34

I am in love with this concept. After getting the Handy and an OSR2+ it immediately occurred to me that if the stroking and twisting and what not could all be controlled how amazing would it be if AI could be utilized somehow to speed up/slow down the stroking automatically in response to the level of pleasure you were currently feeling and be programmed to still finish you after a certain amount of time or amount of edges.

However I wouldn’t even know where to begin to even explore if anything like this would ever be feasible.

Kudos to you on dreaming up the concept, researching how to make it work and implementing it!

This could really be a game changer.

I also agree that services like this and ideas how to better improve various projects shouldn’t always be monetized. I disagree with the poster who gave the 3 reasons he had to charge. While I understand what he is saying, everything they said goes entirely against your idea.

The improvements he mentioned are all for monetary gain. If your goal is as it sounds it is, is to share pleasure with the community you’re obviously not concerned about the money as you’ve stated.

1 Like
35

I randomly found your post via google and I am pretty impressed by the idea and your current progress. Just to think about that it is possible to let a AI control your toy based on my biometric data is just mindblowing.

I can only state my situation: With electronics, I am very bad haha, I soldered only a few times and thats it. I just tried programming a few things in C and… yes.. So I would be one of those who would be thankful, if something like this build can be bought as finished product.
I would also buy the EOM3k, when it would be available in EU, but sadly it isnt…

However I would also be very thankful to hear some updates from your incredible project! :slight_smile:

PS: A honest review about how good the EOM3k or nogasm really works would also be interesting, I only found (I think) one real review with some details about how long it take to figure it out, which settings are right, how it feels, … - maybe someone knows a detailed review about it and can send me a link or so, I dont want to “steal” this thread :wink:

1 Like