/robowaifu/ - DIY Robot Wives

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

Merry Christmas!

Update on the file situation (it's good)

The warrant canary has been updated.

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)


Merry Christmas, /robowaifu/ ! Please join the /christmas/ party this year!


Open file (380.52 KB 512x512 1 (25).png)
Open file (359.76 KB 512x512 1 (58).png)
Open file (360.42 KB 512x512 1 (62).png)
Open file (330.60 KB 512x512 1 (93).png)
Open file (380.42 KB 512x512 1 (104).png)
AI Robowaifu Art and References ** SoaringMoon 11/25/2022 (Fri) 06:43:32 No.17763
I generated a whole much of neat images with Stable Diffusion 1.6. Enjoy. You are free to use the for whatever. >OP images are my five favorite of the bunch. Some proportions are off obviously. < "a robowaifu with [color] hair, digital painting, trending on artstation" Was the generation phrase. --- >Sorry to spoil all your files, rather than just the one (w/ Lynxchan it's all or nothing after the fact). The Problem Glasses are a Leftist dog-whistle that is rather distasteful around here (and also a red-flag). Certainly not something we would want to look at year-after-year in the catalog. Hope you understand, OP. ** Probably best to limit it to image generation, but also tolerating clips. Anything more advanced goes into the current propaganda thread : (>> TBD) . >=== -edit subject -add footnote
Edited last time by Chobitsu on 09/06/2024 (Fri) 09:14:10.
>>32856 >The second one with the number 100 is from the Papercraft Lora. Neat. Maybe we should investigate this further. After all, there's a good reason this approach is used in vidya & also for film blocking-out models. GG, Anon. Cheers.
Open file (138.33 KB 1024x1024 image-116.jpeg)
>>32858 The papercraft variants certainly look interesting, but from a maker perspective I'm more intrigued by the yarn art variants and also all of the ones which have some flexible elements that could be covered over in silicone rubber. Non human skin suits which look sexy are also fascinating. These look better than I would've thought. I'm getting generally more open to some actual robotic looks in waifus, I'm normally the full body silicone skin guy here around.
>>32855 I like the cracked head one image-97 image-113 nice the neck looks more realistic and I like that partial facial sweater. It's odd as can be what these thing spit out. This has nothing to do with robots but it's cool as can be, https://icanbecreative.com/article/if-rome-never-fell-ai-generated-pictures
>>32859 Yeah, it's good. BTW, I'd really like to see the full-body of that panel design by this model. Any chance it can do that for you, NoidoDev?
>>32883 I can try to get something like it, but I don't have the seed for that exact model. As I wrote before, unfortunately it doesn't put that into the exiv of the file or the filename. Hugging Face also has no history. In some cases I saved the seed an tried out different Loras and such.
Open file (135.46 KB 1024x1024 image-148.jpeg)
Open file (150.41 KB 1024x1024 image-147.jpeg)
Open file (131.83 KB 1024x1024 image-149.jpeg)
Open file (102.85 KB 1024x1024 image-146.jpeg)
>>32883 I'm also not sure if it does just make the full body image of the same picture if I would use the same seed. I think what you need is a model that extents a picture, like Photoshop seem to have but I also saw something about Krita getting something like it. The other thing is, all of these test I did before, except the tarot card and anime tarot card in >>32905, I just used the same prompt but a Lora which wasn't made for what I asked for. Normally I should rather change it to "body elements made out of paper" for example. For example picrels are what I got asking the papercraft Lora for a full body, but without stating the body should be made out of paper. Amazing, but not what I asked for. I do save some prompts and try to keep track of some changes, and keeps some seeds for a while, but I can't keep track of everything. Maybe, as much as I am using it, I should switch to the official full version on their website.
Open file (99.19 KB 1024x1024 image-151.jpeg)
Open file (91.15 KB 1024x1024 image-150.jpeg)
>>32906 I thought I've had cracked the code, but it went back to upper body pics. I probably should just look it up somewhere.
Open file (88.14 KB 1024x1024 image-153.jpeg)
Open file (100.46 KB 1024x1024 image-154.jpeg)
>>32907 Got something more full body like. But I'm done with this for now. I can give you the seed and prompt for picrel. Once again, I was reckless enough not saving the seed for the full body one in >>32907, since I thought I cracked it anyways. Also, I'm not sure if you are aware, but these hairs and sheets often look the EVA foam sheets you can buy in many shops. Then again, not sure if these are fire resistant. I once thought of using them as some kind of skin, maybe impregnated with silicone. But they would not be as flexible as pure silicone or fold like textiles impregnated with silicone rubber. I don't see much of a use case for any of this, except for some interesting artistic look. Some Japanese people and weebs are also very much into papercraft. Seed: 3957271607, Lora: Papercraft, 0.95, CFG: 3.5, steps: 34 > a realistic full-body photo of an realistic looking young adult woman, the body has some robotic paper-like shell elements in dark blue, it has no big shoulder pads, the shell is not a suit or dress but part of the body, the lower body is visible down to the boots, doll-like but otherwise realistic looking eyes with a realistic eyeball, a mix of very realistic extremely light blue hair in two tails, realistic eye brows out of human hair, grey blue eye color, both eyes in the same color, very realistic looking textured skin, very smooth but full lips without folds, small chin, high cheekbones, no long neck, realistic look of the photo but really keep the big eyes the picture does not cut off before the feet I can already tell you, increasing the seed for one, only creates the picrel 2, once again not a full body image.
Open file (93.44 KB 1024x1024 image-191.jpeg)
Open file (76.98 KB 1024x1024 image-192.jpeg)
Open file (82.74 KB 1024x1024 image-193.jpeg)
Open file (87.48 KB 1024x1024 image-194.jpeg)
Open file (89.06 KB 1024x1024 image-195.jpeg)
I'm playing a little bit around with values. The HF space I'm using currently fails most times when I increase the steps, which requires more GPU usage. The CFG value, which is about "prompt following" vs creativity, for all I understand. It can also have quite some impact. I played around using one seed, but using different CFG values. I think around 3.5-6 or so might be the best, a relatively high number but not to high (around 10-14) might also be fine. I didn't write the exact values down, but the increase in number below shows a increase in value. Going very low creates a very botched image, I didn't post that image, I can only post five anyways.
>>32908 >>32978 >Got something more full body like. You did it! GG, Anon. They all look great; please keep up the good work. Cheers. :^) <---> BTW, 'alway rember' our rule #1 (spoiler NSFW), NoidoDev. That last one is hot enough that I'm inclined to spoiler it. But as you know, it's all-or-nothing after the fact so I'll leave it this time. But just please be judicious and keep our SFW board status in mind (you can spoiler individual ones on your own at post time). Thanks! :^) >=== -add'l crosslink -add 'spoiler' cmnt -minor edit
Edited last time by Chobitsu on 08/19/2024 (Mon) 23:38:22.
>>32978 These are so cool. You've done some great work on these. I'm saving all of these. Maybe at some point I can use them as references for a starting point cause there's some great stuff here.
Open file (130.50 KB 1024x1024 image-186.jpeg)
Open file (101.24 KB 1024x1024 image-189.jpeg)
Open file (94.18 KB 1024x1024 image-190.jpeg)
Open file (88.95 KB 1024x1024 image-198.jpeg)
Open file (97.61 KB 1024x1024 image-197.jpeg)
>>32980 >>32996 Thanks. I generally have some more of each kind of picture which I didn't post because it wasn't that great or somehow off. Picrels were looking "a bit too cute". With some other prompts I got something that looked like the start of a porn movie, with very human looking women. - Switching between "young adult woman" and "young woman" can make the switch from plastic-toy or cute doll looking and something more mature companion-doll like. But sometimes, it also switches to realistic women without big eyes, and I have to switch back or play with the mentioning of big eyes. Which can be "big/oversized/doll-like but not too big" and so on. - Silicone skin broken into sections is something I never got to work. - Commas and how to phrase things matter somewhat. I think I used this prompt, but it ignored the "from a higher position" so this might be removed, this was also the part that got me quite wilder pics before I made some changes: > a realistic full-body and full scale photo of an realistic looking young adult woman with oversized but not too big eyes while she's kneeling on the floor, the very realistic looking robotic textured pale skin elements made out of silicone in her upper body are broken up into sections, the textured pale skin color of these shell elements is the same as her skin in the face, she has boobs but no nipples, no big shoulder pads, the thighs are visible in the full body picture, doll-like but otherwise realistic looking eyes with a realistic eyeball, a mix of very realistic dark blonde hair in two small tails, realistic eye brows out of human hair, intense green eye color, both eyes in the same color, very smooth but full lips without folds, small chin, high cheekbones, no long neck, functional hands, realistic look of the photo but really keep the big eyes >>32908 Btw, the papercut lora has been removed from where I was using it. No idea about the reason.
>>33004 It's so unbelievable that a computer can do all this with a few key words. Tricky to get them right though. I really like image-197. I also am mostly interested in waifus that look fairly realistic but these shell types are beginning to grow on me. They would save a vast amount of time and labor I would think in building. That last one is so cute. I would that one and the knit girl image-113.jpeg at >>32855 the red head image-46 at >>32818 There's several like her These I like best so far. But there's a lot of good ones. I was going to list a few but there's so many good ones. Ill stop before I fill the page up.
>>32855 >NoidoDev Maybe this will interest you. I have some skin ideas using 3D printers. Maybe with some of the effects of the Yarn girl in the post. image-113 and some of the images recently you posted on the paper girls >>32978 >>33004 If it interest you I wonder what an AI would come up with, being prompted to make the skin of chain mail like NASA's chain mail. Details on the chain mail in the link below in skin section. >>33010
Open file (123.59 KB 1024x1024 image-264.jpeg)
Open file (118.66 KB 1024x1024 image-263.jpeg)
Open file (112.64 KB 1024x1024 image-261.jpeg)
Open file (102.30 KB 1024x1024 image-255.jpeg)
Open file (102.01 KB 1024x1024 image-254.jpeg)
Robowaifu traveling around after (most) humans are gone. Basically like Girls Last Tour. Maybe she travels with her master, though. Pics are your perspective after some apocalypse.
I tested some other image generators. Ideogram and RealVisXL_V3.0_Turbo. Mixed results. Ideogram is partially very realistic, but also not really. It went with too big eyes, but at least it picked up on the prompt. The other one has severe issues to make big eyes and refuses to change that.
>>33029 This brings up an intredasting set of questions, Anon. So, say the Globohomo under the tutelage of their father Satan actually manage to pull this off, and destroy the civilization that men like we Anons have created. Further, say we here (and cadres like us) succeed in our dream of creating an opensauce/smol-businesses-by-the-thousands, where millions upon millions of robowaifus are out there serving their master. Let's make it a Billion Robowaifus functioning before the GH triggers it's destruction on the European-descent nations (just for a nice, even number). Let's say the first nukes/EMPs/whatever take out 99% of the robowaifus from functioning (setting aside similar numbers of Whites/invaders killed too, and entirely-discounting any such attacks against the baste Chinese/baste Russians). So, that leaves about what (please check me on my numbers here) 10'000'000' functioning robowaifus left in our nations? What then? What happens when 10M robowaifus are left wandering aimlessly amidst the rubble of our once-great Western Tradition'' ? While the CY jew Isaac Asimov had his 3 Laws of Robotics, the man Mark Tilden had his own three laws as well : ( >>32792 ) . Tilden's laws were all about the robot(waifu, in this scenario) working to preserve herself in an emergency : > Tilden’s laws of robotics : > A robot must protect its existence at all costs. > A robot must obtain and maintain access to its own power source. > A robot must continually search for better power sources. Presuming that we here win the day in our grand vision, and also presuming that we here instill Tilden's 3 laws into our robowaifus (as opposed to the asimovian version) -- what will the outcome be? How will the 10M robowaifus keep themselves powered up? What sorts of energy systems will survive the fallout (both literal & figurative) of the GH's attacks? Will they themselves be able to rebuild the technology around them (remember, many, many artifacts of tech will survive these events) into a viable power system? Lots of things to think about in such an outcome tbh. :///
Open file (122.01 KB 1024x1024 image-266.jpeg)
Open file (129.14 KB 1024x1024 image-288.jpeg)
Open file (142.15 KB 1024x1024 image-289.jpeg)
Open file (139.85 KB 1024x1024 image-290.jpeg)
Open file (110.16 KB 1024x1024 image-302.jpeg)
>>33069 The reason for my picture generation was more aesthetically motivated, and from a narrative perspective of a possible interesting story for future scifi. But it wasn't thought out very well, I just like the mix of calm post-apocalype and waifus, like Yokohama Shopping Log and Quiet Country Cafe. Also, I didn't make my mind up about humans still being around or not. If not, or not on the outside, then it might be because of oxygen depletion (it mostly comes from the oceans). I don't think your scenario is quite realistic, but we're not in meta here and this is really some edge case aside from storytelling. I'm pretty sure the EMP problem is rather overestimated. I thought about a security mechanism, using heavy and cheap batteries to charge the waifu, and the cable will physically disconnect while she's charging, then it switches back and charges these batteries. Also, I don't think humans would've an energy problem in most scenarios after a big collapse, especially not after deployment of "renewables". The simplest way would be small electro motors and a stream with a dam, I guess. Anyways, in my pictures they still drive around, so they still seem to have some source of energy. I thought about some extreme edge cases when it comes to individual waifus surviving in the wild before. One idea is that robowaifus could've heatpipes in their ribs and heat up water which would power a very small steam engine inside till the batteries are recharged, or as a addition energy boost. This would not just be there for such a scenario but would also charge her sitting around a pool or walking outside in a bikini. > Will they themselves be able to rebuild the technology around them (remember, many, many artifacts of tech will survive these events) into a viable power system? It's up to us to get them there at some point. But generally their purpose should be humans or men in more particular. I once thought about another good trope covering that, where they do survive and use biotechnology and frozen human cells to bring men back. >Picrel I'm a thigh person, how did you know?!
>>33073 >but we're not in meta here Lol, true. See how it's hard to stay on-topic here? :DD (Actually, this convo would be well-suited to +/pol/ funposting zone IMO.) Please feel free to open this up to wider discussion there. I think Tilden's Laws bear fair-scrutiny by us here, of all people. >It's up to us to get them there at some point. I like the way you think, Anon! :^) >we build robowaifus, they save us from disappearing... I almost broached this topic in my initial post of this convo, but though better of it. Certainly a nice scenario for a trilogy of SciFi books!! :) >I'm a thigh person, how did you know?!
>>33069 >Tilden's Laws of Robotics "Protect thy ass, feed thy ass, move thy ass to better real estate." - Dave Hrynkiw with Mark Tilden in Junkbots, Bugbots & Bots on Wheels.
>>33078 Leld. :D
>>33075 >See how it's hard to stay on-topic here? Yes it is. I was going to comment that emp is the least of our problems compared to the "pole shift" that's coming, but didn't. They could run themselves on the heat of bacterial decomposition of vegetable matter by chewing a lot of grass and wood. Maybe they could burn it.
>>33073 Runway video test. It somewhat failed, since in the first video she was supposed to get into the car and drive backwards into the fog. Hmm. Maybe just not a good choice for a prompt. Anyways, the lip sync seems to be nice, considering it's not a portrait. The other one looks not as good as the best of Hedra but better than recent generations. Either way, I'm not going to pay $95 ever month for Runway right now, though I might get a small plan to further test it a few times every month.
>>33117 Very-impressive results, NoidoDev. GG. :^)
Open file (456.86 KB 960x960 lip_sync(1).mp4)
>>33117 Here Pika.art for comparison. It allows only 3-4 seconds of audio, and only from own sources, no generation. I even had to speed up the clip to make anything useful. The videos also seem to not do much, just zooming in and things like that, with some added facial movements like blinking.
>>33139 The video is from Luma again. Didn't follow the prompt, but it's kinda okay. The pics are from Ideogram, which was rather tricky to get anything useful out, since it has a tendency to a certain form of realism and also requires different prompts. It also doesn't follow the prompt all the time, same as Flux but in a different way. I did not ask for these huge eyelashes. >A realistic picture of an doll that has skin like a young woman. She has very huge doll-like but otherwise realistic looking eyes, similar to Alita. The eyes have a realistic eyeball, she has very realistic light blonde hair, realistic but small eye brows and short eyelashes out of human hair, bluegreen eye color and realistic looking skin. She also has very smooth lips, small chin like a female anime character and high cheekbones. Her eyelashes have no extensions. Her neck looks very human like.
>>33166 Next step with Ideogram was to make me some robot bodies. Look really good, tbh. I'm actually getting some likes there. Never happened on any other platform. My pics there will be here, the prompts can be copied: https://ideogram.ai/u/noidodev/pinned
>>33192 All these recent pieces have been interesting, NoidoDev. This specific set reminds me of an idea I had years ago to build some robowaifus with fine, handcrafted wood shells. Certain collectors would go wild for such unique creations. Cheers, Anon. :^)
@Kiwi Do you have any standard you'd like the board to adhere to regarding doll photography? >>33369
>>33383 I think this is the most appropriate thread. If you agree, this thread may as well be renamed to "AI Robowaifu Art and References." This would broaden the scope while keeping the original intent in tact. If SoaringMoon has any objections, I would respect them.
>>33383 Those images are AI generated, not photographs
>>33401 The prompt has no references to ‘doll’ It is a robot That’s why it has wires coming out if it and stuff
>>33386 Concur, wilco. Daily reminder, we actually do have a doll thread here, thanks to Greentext anon. (If you recall the debacle where a 3DPD took & killed /doll/ .) >>33401 >>33402 >Those images are AI generated, not photographs That's not the primary point of my question, Anon. And very-clearly, that generation model was trained on sexdoll photographs.
>>33403 You’re wrong. The Lora was trained on 2004 photographs https://huggingface.co/spaces/multimodalart/LoraTheExplorer
>>33407 OK, granted it's irrelevant to me one way or other. I was simply asking our Mod what he thought about the topic. The question wasn't directed at you. The secondary point being if you or anyone tries to squeeze sexdoll photography onto this board by w/e means, just know it's going to be resisted by me. Fair warning ahead of time. :^)
>>33409 I never posted sexdolls on here so the warning isn’t relevant to me
Open file (541.35 KB 512x768 Ratri.png)
Open file (523.87 KB 512x768 2BAtri.png)
Open file (63.26 KB 512x768 WasWorse.jpg)
Open file (450.61 KB 512x768 NightDateAtri.png)
Painterly, sketch, and lighting are my favorite words to add to prompts. All are made with these 3 LoRA bocchi_style_offset (Improves eyes and face) 2B-08 (More mature structure) ATRItest (Atri) First one has a random Nazrin LoRA for ears, there's heaps of 'em that'll do the same thing. WasWorse was a complete mess, so I quickly scribbled over it to make it barely usable. For personal reasons, I had wanted a more mature Atri. Unfortunately, I could only do so much by combining her with 2B.
>>33477 >Nazrin LoRA for ears Foxgirls! There's a great story on a site of dubious repute 7chan, with such winning deviance as "My Sister is a Dick Girl" but this story is great. About a guy flung into a parallel universe where they have fox girls. It's really good. He, cough, cough, buys one, there that's normal, she is very useful in all sorts of ways, other than what you are thinking. Hiraeth 7ch@n.org/elit/res/26149.html#i26149 In case you wonder MSIADG 7ch@n.org/elit/res/15127.html I think it would be great to have a foxgirl. It might defeat some of the uncanny valley and be seen as more in the vein of cartoon characters. Some foxgirls, https://cdn.akamai.steamstatic.com/steam/apps/2413720/header.jpg https://images-wixmp-ed30a86b8c4ca887773594c2.wixmp.com/f/a44cd4fe-459f-4fd7-987b-e5efc4abad35/da3uh17-6f8f5e61-766b-4982-9820-1464299a7ad0.png?token=eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.eyJpc3MiOiJ1cm46YXBwOjdlMGQxODg5ODIyNjQzNzNhNWYwZDQxNWVhMGQyNmUwIiwic3ViIjoidXJuOmFwcDo3ZTBkMTg4OTgyMjY0MzczYTVmMGQ0MTVlYTBkMjZlMCIsImF1ZCI6WyJ1cm46c2VydmljZTpmaWxlLmRvd25sb2FkIl0sIm9iaiI6W1t7InBhdGgiOiIvZi9hNDRjZDRmZS00NTlmLTRmZDctOTg3Yi1lNWVmYzRhYmFkMzUvZGEzdWgxNy02ZjhmNWU2MS03NjZiLTQ5ODItOTgyMC0xNDY0Mjk5YTdhZDAucG5nIn1dXX0.QWfnJu_-M5CLsrx3tw82J1YJcz2jyHYEaWdx_S8y-Xg https://avatarfiles.alphacoders.com/340/340766.png >=== -obfuscate hotlinks
Edited last time by Chobitsu on 09/28/2024 (Sat) 11:25:07.
>>33797 Lol, I can't leave these as active hotlinks, Grommet. :D >I think it would be great to have a foxgirl. Certainly kemonomimi fans like Gunship would be very appreciative if you created these foxgrill robowaifus you speak of. :^)
>>33798 >Lol, I can't leave these as active hotlinks HAha the MSIADG, well it so perverted, but the Hiraeth one. It's not bad it's a great story. It has some racy moments but it;s a lot more than that. The guy who wrote it is really talented but most of his stories are WAY more...off.
Open file (644.13 KB 512x768 FallFox.png)
>>33797 >Foxgirls! Put a little Senko in my Atri for a foxy Atri! >Buying Foxgirls I'm fully supportive of buying and selling robot girls of all kinds for all reasons. They're cute appliances after all! Those stories made me sad, commodifying and objectifying living girls feels wrong.
Edited last time by Kiwi_ on 09/28/2024 (Sat) 21:35:49.
Open file (156.02 KB 1024x1024 image-344.jpeg)
Open file (204.33 KB 1024x1024 image-347.jpeg)
Open file (160.00 KB 1024x1024 image-343.jpeg)
>>33369 Interesting. Thanks. The first one is hard to distinct from some cosplay or other dress up. But, maybe ask for bigger eyes next time. >>33477 Nice. I never tried myself getting a lot of drawings from AI. The ones I got were mostly or all accidents.
Open file (1.17 MB 1024x1024 00133-3152526083.png)
Open file (1.63 MB 1024x1024 00463-2800305858.png)
Open file (1.95 MB 1024x1024 00092-3552405286.png)
Open file (1.55 MB 1024x1024 00028-2980930835.png)
Open file (1.72 MB 1024x1024 00303-1947355932.png)
I was recently introduced to this model which I call "n40". >https://civitai.com/models/952556/n40naillous-xl It's compatible with A1111, comfy, and forge. Just slap it into your SD model folder and start prompting. Generally speaking, I think it does a better job with hands and prompts involving different characters than Pony does. It still has some issues though in terms of visual slop. Most notably, you want to utilize artist names and styles to really make art pop, otherwise you get more bland outputs I think. I can share some prompts if necessary but mine are nothing to write home about.
Open file (18.51 KB 320x480 maid1.jpeg)
Open file (18.29 KB 474x316 maid2.jpeg)
I found these images when I was searching for images.
>>34664 Thanks. Interesting. But I still think AI to some output that can easily be altered would be better. It should be a workflow of e.g. letting it make such a picture, then export it into a SVG file or so, where one can turn off the colors and add the back later after making changes to the elements. Idk, this here is nice for pictures, but we really need clean animations with consistent characters.
Open file (1.60 MB 1024x1024 00082-2504567657.png)
>>34787 I would agree, it should be seen as more of a tool akin to photoshop or krita (funnily enough krita recently added some AI tools but I haven't had time to look into them). I'm optimistic as said tools on the market are relatively new. In the next decade, you'll have AI's that can pump out 3D models that can be 3D printed, Animations, and more. All of which will be customizable via an easy to understand interface. It may very well get to the point that you have kids pumping out feature length productions. There's already some of that right now on youtube but most of that AI content is rather sloppy. Some folks are pushing it in interesting directions though. >https://www.youtube.com/watch?v=kqwm7rdcX9E Then you have AI asisted frame painting with shit like this which is also neat. >https://www.youtube.com/watch?v=PugcxQYJGjs&t=3s

Report/Delete/Moderation Forms
Delete
Report