/robowaifu/ - DIY Robot Wives

Advancing robotics to a point where anime catgrill meidos in tiny miniskirts are a reality.

Build Back Better

More updates on the way. -r

Max message length: 6144

Drag files to upload or
click here to select them

Maximum 5 files / Maximum size: 20.00 MB

More

(used to delete files and postings)


Have a nice day, Anon!


Open file (2.07 MB 4032x2268 20220520_071840.jpg)
Ashiel - A Robowaifu Design Project SoaringMoon 05/20/2022 (Fri) 11:22:02 No.16319
< Introduction to This Thread This thread is going to be dedicated to my ongoing robowaifu project. This isn't exactly new, I have mentioned it here before in passing. However, this is the first thread I have opened specific to my robowaifu project and not an artificially intelligent chatbot. This thread will be updated when I have something to say about robowaifu design, or have an update on the project. Most of the content will be of the kind of me proposing an idea or suggestion for developers to make the construction of a robowaifu easier. My design philosophy is one of simplicity and the solving of problems instead of jumping to the most accurate simulacrum of the female human form. Small steps make incremental progress, which is something the community need because little progress is made at all. What progress we do make takes years of work, typically from a single person. Honestly, I'm getting tired of that being the norm in the robowaifu community. I'm frankly just done with that stagnation. Join me on my journey, or get left behind. < About Ashiel ASHIEL is an acronym standing for /A/rtificial /S/hell-/H/oused /I/ntelligence and /E/mulated /L/ogic. Artificial, created by man. Shell-Housed, completely enclosed. Intelligence and Emulated Logic, are both a combination of machine learning-based natural language processing and tree-based lookup techniques. ASHIEL is simply Ashiel in any future context, as that will be her name. Ashiel is an artificially intelligent gynoid intended to specialize in precise movement, and engage in basic conversation. Its conversational awareness would be at least equal to that of Replika, but with no chat filtering and a much larger memory sample size. If you want to know what this feels like, play AIDungeon 2. With tree-based lookup, it should be able to perform any of the basic tasks Siri or Alexa can perform. Looking up definitions to words over the internet, managing a calendar, setting an alarm, playing music on demand... etc. The limitations of the robot are extensive. Example limitations include but are not limited to: the speaker will be located in the head mouth area but will obviously come from an ill-resonating speaker cavity; the mouth will likely not move at all, if so not in any meaningful way; The goals of the project include: basic life utility; accurate contextual movement; the ability to self-clean; ample battery life; charging from a home power supply with no additional modifications; large memory-based storage with the ability to process and train in downtime; and yes, one should be able to fuck it. This is meant to be the first iteration in a series of progressively more involved recreational android projects. It is very unlikely the first iteration will ever be completed of course. Like many before me, I will almost certainly fail. However, I will do what I can to provide as much information as I can so my successors can take up the challenge more knowledgeably. < About Me I am SoaringMoon, a tabletop game designer and artificial intelligence safety researcher. I consider robowaifu a hobby, and a subject in the interest of curiosity. I am an advocate for robosexuality and have always been attracted to androids and gynoid robots. I have no idea why. It is nothing to do with fetishism, sexual gratification, or anything of that sort. I just find artificial feminine forms attractive, and see potential in the clinical precision of raw information-derived conversations from chatbots. I can't wait to see what the future holds there. I have an intermediate amount of 3D modeling experience, but no experience in 3D animation or rigging. I do however have experience in building evolution simulations containing creatures with basic neural networks. Lastly, and most controversially, I am vocal about not being afraid of what other people think of me or my personal vocational interests. I realize I am dangerously easy to find, and there are people willing to try to "cancel" me just because I want to build a fuckable female android. I don't care if you are a chauvinist or feminist, left or right, straight or gay, vaxxed or unvaxxed... all I care about is that you support people's creative efforts or attempts to share knowledge with others. It benefits every single person regardless of your personal opinion. It doesn't matter what you think about companionship robots, robot secretaries, literal feminine robot dishwashers and sandwich makers, or sexbots. For every alleged chauvinist who wants a female robot sex slave, there are 10 elderly people who can have a full-time conversational companion capable of doing the chores they are physically unable to perform. Allowing them to live a comfortable end of life without the need of a nursing home, or hospice care. If you are against that you can fuck right off. < Image: ELENOR (Phantasy Star Online EP 1)
Glad to see your project SoaringMoon! Look forward to seeing good things with it. Please feel free to contribute suggestions or other help to various ongoing projects here when you can do so. Cheers.
Open file (308.13 KB 1920x907 kakar-cheung-1.jpg)
Open file (1.75 MB 1920x1920 kakar-cheung-2.jpg)
Open file (241.42 KB 1920x870 julia-urchenko-3.jpg)
>>16320 Thank you, will do. I have been for a long time, just without the trip. < Kakar Cheung and Leg Design For my first design topic, let me discuss leg design and optimization for android frames. Because of the obvious bipedal balance issue, it is important for robowaifu designs to be bottom-heavy. Its taken several thousand years of evolution to get the human brain to balance bipedally for any significant period of time. It's time to talk about the other animals able to stand on two legs... birds. Birds are able to balance the way they are able to due to the fact that what we see as their knees, are actually their ankles. These long feet create a "reverse bend" or "digitigrade" which "increases the economy of running by lengthening the leg and improving the storage and recovery of energy in the tendons and ligaments of the lower limb" (National Science Foundation Feb 2017). It also has the advantage of better carrying and absorbing stress from the weight above it. Unfortunately, this doesn't help us in creating an approximation of the female form; as humans do not have a digitigrade leg structure. So we have to come up with another solution. Let me introduce to you the art of Kakar Cheung, depicted in the post. > https://www.artstation.com/liluuu > https://julia_arty.artstation.com/projects/3z5qD (Pink 3D Girl, collab with Kakar Cheung) His stylistic signature for the female form accentuates the legs (and arms). He widens, lengthens, and blocks out the legs much like one would when sketch drafting. Another art style similar to this, albeit much less refined, is that of the anime *Panty & Stocking with Garterbelt*. which has a reverse taper all the way to a very wide foot which would be impractical for moving about in human-occupied space. For this reason, Kakar's leg proportions are almost ideal. These thick polygonal legs of nearly uniform thickness provide many advantages. The greatest of them is that they contain an abundance of space for mass storage. Keeping the mass at the bottom of the skeletal structure helps the servos move the weight upward and across the ground. Instead of having to fight against the forces of gravity moving the top of the robot around; far fewer adjustments need to be made to keep the robot upright. If planned properly, the robot can even be self-righting. This mass doesn't have to be dead weight at all mind you. Do you know what component in a robot is the heaviest? Something that it needs alot of? That's right, batteries. Putting all of the battery power into the legs solves so many of the balance and power access issues, that it's basically like cheating. All of which can be done without compromising the femininity of the android. Thick thighs, waist, and hips; Long legs, short torso, high leg to torso length ratio. These are all highly attractive features of the female form. The blocky low poly nature of these legs means easy fabrication and construction. Of course, these exact angular features aren't necessary. A "rounded" version of this kind of leg would be better suited for appearance. The only major downside is obviously its bottom heaviness. Lifting or moving the android unpowered would be a deadlift every time. Moving, even with a dolly, pallet jack, or other mechanical assistance would also be difficult, as the lighter top could drop or slouch. Moving it would be like moving a heavy two bungee-cabled CRT televisions with a modern large flatscreen tied between them. Not fun in the slightest. Even when self-powered, any action that would result in the robot being on top of you would be heavily biased to the legs. It's amazing such a robot kneeling on your arm, or propping its legs on you anywhere.
>>16321 >It's amazing such a robot kneeling on your arm, or propping its legs on you anywhere. Just imagine such a robot kneeling on your arm, or propping its legs on you anywhere. Grammarly did that I promise. :P
>>16321 I agree about the character design appeal of the 'chunky' legs. In fact MaidCom, Pandora, and Reploid projects are all tending towards something similar in form. While I certainly understand the concepts behind making a robowaifu 'feet-heavy' there are challenges associated with that approach, some of which you mentioned. The primary issue with this basic approach IMO comes during later revisions of robowaifus where we'll be going for more realism and life-likeness. Human, bi-pedal walking can actually be likened to 'controlled-falling'. That is, we are pretty always just right on the edge of toppling over during normal walking. God engineered us just so, since this is actually the most highly-efficient state (energy-wise) for us to maintain during bipedal kinematics. As an amateur boulderer (rock-climbing) I can assure you that it's our bones, not our muscles primarily, that keep us upright. You can rest for many minutes if you lock you bones in place. They will support you! The issue for us as engineers with the audacity to try to reproduce the design of humans, is this very instability itself. We're far more like an twitchy racing machine in our bodies than we are like a lumbering cargo truck. The first is unstable, yet fast. The second won't win any races, but generally it's much more stable. Our trade-offs and challenges to come are somewhat like that. We need to devise Inverted Pendulum systems that are always right on the verge of falling over if we hope to ever reproduce accurate facsimiles to typical human walking/running gaits. Good points Anon! Thanks for your thoughtful effort posts. That is much-appreciated here. :^)
Also >obligatory crosslink (>>237 - Bipedal Robot Locomotion General)
>>16323 Thankfully, for we have the advantage of using space age materials over bone. Bones self repair, titanium alloys ideally never have to be repaired. Bone is limited by what it has to do; support the weight of a structure while remaining pliable to impacts, sudden shocks, and vibrations. Androids don't have have to support the limitations of complex biology. We can just use the strongest, flexible, composite of materials and structures. We don't have to worry about blood vessels, nervous systems, or muscle tissues. Honestly, we don't even have to try to emulate them. We only have to make an "passible" simulation of them. They only have to feel and look like they do, what they actually do internally doesn't matter. As long as the result is approximately human.
>>16319 Welcome SoaringMoon, hope to see great things from you, a fellow like mind. Have no fear, we are all here to clang and build helpful companions. No one native here will call you a chauvanist or attempt to cancel you. >>16321 >>16323 Chobitsu beat me to the punch. Though I would add that putting the batteries in her feet to act as swing mass when pendulum walking like a human can be a great idea if the necessary timing is accounted for. It does require attention to how god built us, as Chobitsu mentions. A mix of Cornell's Ranger and Femisapien seems to be your best bet SoaringMoon, hope you'll make your project compatible with MaidCom. http://ruina.tam.cornell.edu/research/topics/locomotion_and_robotics/ranger/Ranger2011/ >>16325 >Titanium >Designing to not be repaired and not for repairability. As someone who's been at this for years, repairability and using easily sourced and used materials is important.
>>16326 >Designing to not be repaired and not for reparability. Oh you misunderstand, I'm not against reparability. (In fact I advocate for Right to Repair.) I mean, you should design with materials that ideally shouldn't require repair to begin with. Solid poles of metal that do not lose ductility due to work hardening, because they are not flexing. Sure they may break eventually, but it is best to ensure they aren't a point of failure.
Open file (87.61 KB 846x892 current_level.png)
>>16327 If anyone wants to know how well the chatbot is, this is what conversation was like while training it today.
Open file (120.47 KB 933x1252 FOMGu0fX0AIMQLn.jpeg)
Open file (471.49 KB 869x491 giantsumomo.png)
>>16321 I like this idea but I think this style will only look good up to 1/4 scale which will make it difficult to put any servos in the body or arms. The style would have to be adjusted to accommodate parts while keeping the appeal for the desired scale. Stuff that looks cute doll size looks really funny scaled up. Also putting batteries in the legs is a huge safety risk. Robots are prone to bumping into things and repeated mechanical stress on the battery will wear out its separator and eventually cause it to catch fire or explode. It's not rare for RC cars and drones to burst into flames after landing hard or excessive crashing. I've had a small battery explode and shoot shrapnel across my room just sitting unplugged on a shelf a few days after dropping it. The thought of a larger battery exploding is fucking terrifying. The main reason robots are difficult to balance is because they lack a spine that can bend in any direction to quickly correct an imbalance. Balance is an easy to solve problem so long as there is a way to solve it. Most robot designs are stiff as boards though so once they start to tip there's no physical way for them to bend and shift the center of gravity to counter the fall.
Open file (554.40 KB 869x491 scale.png)
>>16355 I disagree, what I suggested doesn't look too out of place at full scale (albeit the head is a little big). >I think this style will only look good up to 1/4 scale The images depicts a 1/4th scale (to scale lol). Honestly that is a bit of an exaggeration, but what you are suggesting "only looks good at" is actually quite small. >Also putting batteries in the legs is a huge safety risk. Not any more a safety risk than an electric vehicle battery with the shocks of road travel. >I've had a small battery explode and shoot shrapnel across my room just sitting unplugged on a shelf a few days after dropping it. Your anecdote is a little suspect. I'm sure it happened to you in the way you've described, but I've never seen a video recording of a "battery detonating" in any significant manor. You'd have to show me an example. < This is what large battery explosions look like. They are basically dangerous fires. https://www.youtube.com/watch?v=Qzt9RZ0FQyM But again, no more so than an electric vehicle. Your robot has to contain a large battery, there is not a compromise. You cannot have the android without one, it just can't be done. > Difficult to balance is because they lack a spine that can bend in any direction to quickly correct an imbalance. When it comes to the balance of a machine, shape and range of motion has little to do with it. You can always use flywheel stabilization gyroscopically. These can be spun up on demand before righting and do not need to be run constantly. I'm not saying this is the desired solution, I'm saying this is an example that does not require a spinal linkage. Of course having a spinal linkage has alot of advantages.
>>16325 >Thankfully, for we have the advantage of using space age materials over bone. Indeed. I'm certainly on the same page with you there Anon. (>>16059). OTOH, we are well-served by bio-mimicry in general. Cf. this "Byrun" video, for example (>>14578). >>16327 >In fact I advocate for Right to Repair. I consider this non-negotiable for anything produced by /robowaifu/ tbh. >tl;dr We will never intentionally put out something that Anon can't hack+mod at will, as long as I have anything to say in the matter. >>16359 >The images depicts a 1/4th scale (to scale lol). Lol. Obvs, dear Sumomo-chan is intended for fairy-bot scale headpats! :^) === Comment: Regarding so-called 'Right to Repair', we obviously cannot make any stipulations regarding components we ourselves aren't manufacturing. For instance, can Anon hack+mod a RaspberryPi? Probably not much, in general. >=== -add 'Comment' cmnt clarifications -minor grammar edit
Edited last time by Chobitsu on 05/24/2022 (Tue) 05:52:57.
>>16326 >Cornell's Ranger Neat. Yep, 'stilt-walking' very much conveys my thinking about the efficiencies involved with the IRL bio-kinematics of a bio-tensegrity system using the human skeletal system. It's all quite amazing (humans, I mean) the deeper you dig into it. >>16355 >and repeated mechanical stress on the battery will wear out its separator and eventually cause it to catch fire or explode. Another reason to keep batteries near a robowaifu's central-mass point (and preferably protected inside some type of impact-resistant cage, similar to the robowaifu's breadbox). Great point about the spine Anon! You might read up the bio-tensegrity/meshtegrity discussions in the MaidCom thread (>>16057, ...) .
>>16359 LiFePO4 is definitely the best possible option for waifu's. If using that chemistry, it's mostly safe to keep it in the legs. Though, as has been mentioned, othe Li based chemistries would be dangerous in the arms and legs. >Balance As someone who did much research into balance years ago; shifting upper body mass or a low center of mass have many benefits compared to reaction wheels for bipeds. These include, much lower energy consumption with low center of mass designs essentially requiring no energy for balance, though more energy to move. Simpler design, and they're easier to program as reaction wheels are finicky to get right. You could also just give them a tail, the easiest way to attain balance is to increase points of surface contact. A dress with wheels also helps.
>>16404 >Lol. Obvs, dear Sumomo-chan is intended for fairy-bot scale headpats! :^) Well she is the size of a christmas ornament lol. >Right to Repair' I would stay away from any astroturfing buzzwords of the day. That term implies we sell things here and can sue them for fixing or modifying the product they bought.
>>16426 >astroturfing buzzwords No, Right to Repair implies we don't sue them for fixing or modifying. Right to Repair is allowing side, not the denying side. >implies we sell things here Yes, did you think if I'm building an android for private use that you would be getting it free of charge? Advocating for right to repair means that if I do sell it to you that you would have permission to modify the product you own. The product should come with a manual that includes electrical and engineering schematics to better facilitating you do that. Calling it an "astroturfing buzzword" undermines the entire ideology of hacky customization of one's own electronics.
>>16424 One thing I'd like to point out Kywy regarding my regular encouragement to 'centralize' robowaifu's mass around her center of gravity is with longer-term, fully-bipedal-capable robowaifu designs in mind. Basic physics for our more advanced mechanical designs to come, yeah? But please understand I'm not in any way suggesting the approach we're adopting in MaidCom initial prototype is somehow flawed. Indeed I think it's literally the best compromise solution we can manage ATM for something that just werks + provides the basic facilities you've set for the project's goals. >tl;dr Drive on! :^)
>>16426 >Well she is the size of a christmas ornament lol. Heh. Headpat Waifus is the new category I intend to use for the union set of Fairy-bot + Moe-bot robowaifus. :^) (>>13018, R/W form factors related) >I would stay away from any astroturfing buzzwords of the day. Understood. The abuses of Apple and other lawyer-filled mega-Globohomo corporations sparked this debate long ago, it's nothing new. And actually, it's a real & legitimate grassroots resistance to some of the Globohomo's evil shenanigans, AllieDev.
Open file (6.73 MB 500x800 0000.gif)
This is the first time I've ever modeled a humanoid.
>>16589 >This is the first time I've ever modeled a humanoid. Neat, nice beginning Anon! So, it turns out that studying real-life anatomy and making studies & sketches is a key to becoming a good 3D modeler, who knew? You might try doing some life-drawings and even from just reference pics of human beings & animals if this is something you find to be interesting. I'd suggest also, that you just use a traditional, slow-rotation 360' 'turntable' orbit for your display renders. Helps the viewer get a steady look at the model right? Good luck with your efforts SoaringMoon! Cheers.
>>16589 Looking pretty good SoaringMoon!I like the low poly aesthetic. Are those orbs planned for use as a mating feature?
>>16621 Kek, forgot mating had other connotations. By the way, what are you using for modeling?
Open file (7.70 MB 3091x2810 waifu_edit_4.png)
Open file (2.19 MB 1920x1080 ashiel_wallpaper.png)
>>16624 Just blender. Let me post my stuff from WaiEye here as well if anyone wants to use them for whatever reason. >I've been having a field day with VHS effects after learning how to do it.

Report/Delete/Moderation Forms
Delete
Report