>>38405 Not sure about C++ facial animation. Right now I'm just using KDtalker with a lower frame rate. It takes ~30s to get 15s response with piper TTS on a 3090, so like 5 more seconds for full voice cloning. It only works for humanoid avatars though. https://github.com/chaolongy/KDTalker You just have to change the inference.py to the audio\image file of your bot program. That gives you a full video\audio cloning head avatar. Full body with prompt direction is close with framepack. That would work with non-humanoids too.

>>38405 >The CCP might spy on me, but that's fine. I just won't talk to it about tiananmen square >Lol, good advice. My own personal take on this (being in Burgerstani r/n) is that I'm far-less concerned what Dr. Lee or Mr. Wong think of me, and far-more concerned about some mentally-deranged glownigger sicking some local yokels on me for my doubleplus-wrongthing; ie, the so-called Troon-Thug Cycle. ** >That could actually affect my life. What the Chinese do or don't think about me is of little concern. I 100% agree. Every time someone brings up spying through Chinese technology, I always say "what's Xi Jinping going to do with my data?" Better someone far away and removed than people at home. Not to sound like a CCP shill, but westerners have some nerve pointing the finger about spying and political censorship. A Russian and an American get on a plane in Moscow and get to talking. The Russian says he works for the Kremlin and he's on his way to go learn American propaganda techniques. "What American propaganda techniques?" asks the American. "Exactly," the Russian replies.

>>38410 OK, understood Barf. I'll just munge something together as suits me along the way then. Probably around the start of Summer would be my current estimate. Stay tuned. >KDTalker Neat! Looks pretty sophisticated. I'd love to be able to take a whack at this myself sometime. Cheers, Anon.

>>38411 >I always say "what's Xi Jinping going to do with my data?" Better someone far away and removed than people at home. >ur pics-related Exactly. The GH-kikes are hoping that the old Cold-War era brainwashing on their parts is still in play -- rent free -- inside the heads of Westerners today. LOL They somehow seemed to forget -- we don't consoom their Idiot Box day & night! >Not to sound like a CCP shill, but westerners have some nerve pointing the finger about spying and political censorship. This. And as to the CCP, I'll totally shill for them instead! :D In a strange turn of bizarro-world events, the baste Chinese -- under Emperor Xi's tutelage, no doubt -- have in-effect become a National Socialist country today (or close enough). Meanwhile it's the w*st that is the den & haven of (((Filthy Commies)))... both within the governments & without! >tl;dr What a strange timeline this is, Anon. That those silly Chinee will be the ones who usher the world into the Robowaifu Age!! Buckle your seatbelt, Dorothy... <---> >A Russian and an American get on a plane in Moscow and get to talking. >The Russian says he works for the Kremlin and he's on his way to go learn American propaganda techniques. >"What American propaganda techniques?" asks the American. >"Exactly," the Russian replies. KEK Perestroika/10 - would Glasnost again

>>38419 >DA GOVERNMENT IS LE SPYING ON ME THROUGH MY OPENAI KEY >DEEPSEEK IS OK THOUGH, IT IS JUST IS, OK??? Just say you goon to loli all day lil bro it's not that hard

>>38430 Kek. Clearly you're not following the play-by-play here, fren tourist. Welcome! Please have a look around the board while you're here. :D

>>38433 Based on them making fun of our Corpo API privacy concerns and our preferences of DeepSeek, I suspect it's someone from CHAG

>>38434 Heh, maybe, maybe-not. Regardless, libshites/filthy-commies/kikes/glowniggers have been a regular here since ye olden times. Oddly-enough some of them recognize the fundamental existential crisis that robowaifus mean to their scams against humanity, while many Anons did not (though I think that's slowly changing). >tl;dr Meh. The redpill is too hard to swallow for them, let 'em rot in it if they insist ('you can lead a horse to water' & watnot). We'll still press on to achieving this age-old dream, all the same. Cheers, Anon. :^) FORWARD

>>38436 I did consider that, but the specific mocking of our disdain and arguments against Corpo AI keys reminded me of that thread.

>>38437 True enough, GreerTech. OTOH, if even just one of their honest cadres comes over to recognize the flaw in their reasoning on this topic, then I'd consider that a win for us here. And its not that hard a get for them either; after all, /lmg/ has been a thing on 4cuck since the original LLaMA """leak""". :^)

Funny seeing all the factions fight. The same factions probably exist in the 3-letter agencies, as we've seen their chat logs. Just imagine what they have since xkeyscore over a decade ago. If your local fusion center doesn't like you because of a bumper sticker, you are now a target for harassment. It goes both ways though, so we'll see how that all plays out. As for visual avatars, there will be corpo models. Elon will probably make a catgirl avatar for grok, zuck his metaverse version, microsoft will make clippy and that will get someone off no doubt. All of those will have to comply with subpoenas, and then it's just who is doing the enforcement. I really dont care as a lolbert. I think some welcome the authoritarianism as long as it is for their tribe. I don't have a tribe - hence robowaifu. I'll just build my own roads and robots. It's not that hard if you have the capital.

>>38439 >Elon will probably make a catgirl avatar for grok Kek'd . I'm still holding out hopes he really meant that, and wasn't merely sh*teposting! :D >It's not that hard if you have the capital. Trust me bro, if I already had that kind of capital, then the whole world that wanted them would already have -- here, today -- freely-available, pleasing & effective, opensource (MIT-licensed; both h/w & s/w), 100% offlined, locally-run robowaifus+kits!! This is what we're striving for here, and nothing less will truly suffice. Cheers, Barf. :^)

>>38440 >100% offlined, locally-run robowaifus+kits!! That's the dream. I'm just not sure they should be subsidized or we'll get a one size fits all robowaifu. But, I support the concept and look forward to future iterations.

>>38443 >I'm just not sure they should be subsidized or we'll get a one size fits all robowaifu. It's a fair point. As counterpoint, I offer the Wright Brothers and their flying wonder. Until their breakthrough achievement, there were no powered flying machines. Once the breakthrough occurred, then men from nations all over the West were inspired and in less than 20years later, there were literally dozens of flying machines in production. I consider this kind of like 'breaking the 4-minute mile'. Somebody's gotta do it first...may as well be us here on /robowaifu/ . >tl;dr Till now, there are no widespread robowaifu designs, suitable as kits for everyman. I'm pursuing the business acumen to achieve the funding to start a company doing just that. Until then, Anons can only dream of the future. I personally don't consider that enough! Simple as. :^) TWAGMI

>>38444 >a new craze a la wright brothers At the risk of derailing the thread, a major issue is that any dork who rubs two wires together thinks they're Tony Stark, and create videos for ad revenue rather than create a product to sell. Something like this caused the Mech Crash of 2018: a bunch of companies built "mechs" in response to the "Megabots vs Kuratas" hype train until the fight actually happened then disappeared due to all the negative press (Megabots were chill but kinda fly-by-night while Suidobashi Heavy Industries were stuck-up pricks). I can't help but feel we're seeing a psychohistoric echo with all these "advanced humanoid robot" videos on youtube. I heard there was a chinese robot foot race that got heavily censored because most of the robots would take 2 steps and fall over. Unless the robots are out in random crowds (like elon did with having optimus being a telepresense device) or actually available for purchase all "advancements" should be taken with skepticism. After all, it was revealed one of Amazon's "Smart Stores" wasn't run by AI but a bunch of Indians watching via camera and taking notes... >>38443 >That's the dream. I'll try to get around to some durability testing for Pringle at some point, because if she passes the durability tests she'd be ready for public release and I made sure that she is very easy to work with :)

>>38444 Not to toot my own horn, but I feel like Galatea is that kit. It's finished and published, and available for any to build. Now my hope is that it will be the Wright Brothers moment >>38443 >That's the dream. I'm just not sure they should be subsidized or we'll get a one size fits all robowaifu. But, I support the concept and look forward to future iterations. Agreed on the design philosophy. I don't think there will or should be one "The Robowaifu, alright pack it up everyone". Multiple people have different ideas, desires, and visions.

>>38453 >>38454 LOL. OK, ok guys -- you've convinced us... THE FUTURE IS HERE

>>38460 I probably just watch too much animu, right? :D "Chii?"

>>38453 >she'd be ready for public release and I made sure that she is very easy to work with :) Nice! She'll be a great part of the first wave of real robowaifus!

So, how do we go about promoting & distributing the robowaifus that are already here? My mental canon involves a self-supporting business doing that (for-profit sales [fully-assembled or kits], freely-distributed designs, software, & kit-assembly support). Since that business isn't here yet, how can we achieve something similar without it?

Obviously, we are all derailing the Visual Waifus thread r/n. Let's move to the thread of your choice please, Anons.

>>38454 >>38453 I think Galatea could be sold on Etsy or where ever as is. Hope to see you guys sell kits. Maybe just a marketplace thread with links to different projects. I was mostly joking about the subsidy thing. If my tribe's casino produced enough money to hand out robowaifus to tribal members, I just fear they would also look like my tribal members, so I'd go for the Chinese silicone version and pay for it myself. I mean, I wouldn't refuse the state mandated robowaifu of course. I'm just not sure I could move the thing. For visual Chinese made waifu, here's one I have on order right now. It has moving hips, neck, tongue and built in AI. Not bad for $2k.

>>38464 Yup! One thing about virtual waifus (and to circle back to my potentially derailing post) is that due to their nature it is pretty easy to prove if the maker is lying about its functionality. Eg "want to try out my virtual waifu here's the download link/place to buy" vs "I built a virtual waifu but no I won't share/monetize it". No need for specialized hardware a lot of the time. I have a similar program that has little SOS brigade members run around my screen with some "Disappearance of Haruhi Suzumiya" action. The program made in like 2009.

>>38453 I still believe in the heart of the cards mecha. I think right now the mech industry is going through its equivalent of the period between the first video game crash and Nintendo reviving the industry. Somebody will succeed in making mecha popular and profitable enough to build a business on at some point. I think there should be a thread about it, but I tried last year to make a thread about covert ops robowaifus and apparently that's considered off-topic here, so I don't know how you could justify having an /m/ thread here. Also even if there was a thread I don't know that I could be around enough to participate meaningfully in it. I've always lamented not being able to do more to further the robowaifu cause, which is partly because I'm busy with other things and partly because I have little in the way of skills that would be useful for this.

>>38474 >One thing about virtual waifus (and to circle back to my potentially derailing post) is that due to their nature it is pretty easy to prove if the maker is lying about its functionality. Good point, Mechnomancer. >little SOS brigade members run around my screen with some "Disappearance of Haruhi Suzumiya" action. A cute! Giving them contextual understanding now, and the ability to TTS<->STT would really liven things up! Cheers, Anon.

>>38476 >but I tried last year to make a thread about covert ops robowaifus and apparently that's considered off-topic here Self-defense/perimeter-patrol would certainly be welcomed and on-topic in : ( >>10000 ). However, the last thing I want /robowaifu/ to become is "Terminators-R-Us". While your effort is appreciated Anon, your topic is far too close to that destination. My apologies again, and I wish you good luck on that idea in another forum! >I've always lamented not being able to do more to further the robowaifu cause, which is partly because I'm busy with other things and partly because I have little in the way of skills that would be useful for this. Ehh, many of us are busy, Anon. I think just participating by posting in threads here on the board is sufficient to at least be an encouragement to everyone! Please continue doing so, Anon. Cheers. :^)

>>38479 Meh, it's no big deal. I've had threads closed for much dumber reasons than that. But where's the line between teaching robowaifus self-defense and teaching them to be soldiers and assassins? And also, what can somebody who doesn't have enough programming or engineering skill to be useful in building waifus do to contribute?

>>38480 Also not sure where that line is. Legally, I think most places wouldn't even allow them to have non-lethal options like pepper spray. But back to vtubers. I know nothing about them really. Has anyone else got this working and familiar with Live2D? I'm just using the default characters and there isn't a good english site for them https://docs.llmvtuber.com/en/docs/quick-start/ https://docs.llmvtuber.com/en/docs/user-guide/live2d Looks like open vtuber uses an older version of Live2D so a lot of the models won't work. Also hard to rig non-anime characters\robots. Installing open vtuber is easy if you already have CUDA\ffmpeg installed. I just had to install UV and make sure it is in my PATH statement. After that, it was just 2 commands "uv sync" and then "uv run run_server.py" to run it. Just have to edit the conf.yaml for your preferences.

>>38476 >Somebody will succeed in making mecha popular and profitable enough to build a business on at some point. Me. I originally started robowaifu development an intentional byproduct of investigating computer controls for my mech projects (I mention these a bit in my first thread about SPUD). They are pretty much like smaller robotics projects, the only difference is I'm using parts that nobody really thinks about using. And in this manner I create constructs at a fraction of a percent of the cost of competitors. It might be fun to make an AI vtuber version of my mech. I've been working with the Pillow python library and its ability to composite images so making an entirely python-based vtuber system might be a possibility with inputs to the avatar determined via a dropbox file. IDK, a little bit outside my current scope.

>>38486 >Somebody will succeed in making mecha popular and profitable enough to build a business on at some point. >Me. That goes hard

>>38486 >Me. This.

>>38486 >Me. A bold claim. It's good to have people who are willing to shoot for these big goals, but how do you intend to get the startup capital needed for this? Although if you can make good on this, you'd have ample money to pour into robowaifu development.

>>38498 > how do you intend to get the startup capital needed for this? 2nd pic is my mech family as of '22. Been going to local events since '20. Carry the workshop waifu is the first one to have a computer brain. I'm still working on making the others computerized.

have you taken notice of that super popular project on kickstarter? I got an ad for it recently. https://www.youtube.com/watch?v=WFgXunR8b6A&ab_channel=gizmochina

>>38612 Thanks, Anon! That's very interesting. I wish them good success with that product line. Cheers. :^)

>>38612 At first I was skeptical because it looked like a standard corpo AI waifu, but I checked out the kickstarter page right now, and it does support offline AI and custom models. So I think it's going to be pretty good! Exciting times we live in. The first generation of AI waifus and robowaifus are coming out.

>>38615 >Exciting times we live in. The first generation of AI waifus and robowaifus are coming out. This. What a time to be alive!! FORWARD.

(Edited from Trashchan) Introducing Galatana, the standalone AI system. It uses the same AI used in Galatea v3.0.4 >>38070 Perfect for more budget oriented anons, or anyone who doesn't want to or can't build a full robot. You can talk with her anywhere, by using a single bluetooth earpiece and your phone https://greertech.neocities.org/galatana ------- Original Trashchan post https://trashchan.xyz/robowaifu/thread/84.html#bottom Odysee Backup https://odysee.com/@GreerTechother:3/Galatana:4?r=2RAnQD4k7R6nPoYVC32GJaXWCqK6sKmT

>>38897 Ambitious project! I thought about making a robot as well for the last week. I asked the guys at work who do point clouds about that and I think for now it would be easier to do holographic projection with AR. I downloaded https://github.com/PKU-VCL-3DV/SLAM3R and converted a 30s video of my house into a pointcloud. There are other projects to convert them into 3d meshes and also label them as household objects. Open3D TSDF and Meta Segment Anything Model 2 (SAM 2). At first I thought of making a pipeline that updates the meshes and segmentations in real time for a robotic agent to navigate my house. Then it hit me, why not use the internal representation to create a virtual environment of my house and let an AI agent navigate that? AR can be used to display them. Pathfinding is possible with the meshes, and with labeled objects I can create a graph with the relations between objects. The AI agent navigates this in some 3d engine. With simplified meshes that match my house layout. And send the agent mesh to a separate render target on my AR headset.

>>39621 Thank you :) That point cloud creator is a really good find. Much better than most video-to-point cloud processers I've seen. >At first I thought of making a pipeline that updates the meshes and segmentations in real time for a robotic agent to navigate my house. A lot of high-end robot vacuums use LIDAR for a similar purpose, and a 3D point cloud will definitely help for humanoid robots to navigate, what may be unobstructed to the base may be an obstruction for the whole robot (ex. a table or a chair) >AR hologram waifu I like your idea. You're clearly very knowledgeable and talented. (second pic related)

>>39624 I tired out some projects like vox fusion and another one with ros. Couldn't get either of them working. Vox fusion had a demo unit that just didn't work and I can't be arsed to install linux for ros. I tried installing PopOs like 2 weeks ago for some triton project. It just didn't work. I spent 8h getting cuda and gcc to play nice but no success. I was yelling at my pc at this point. I gave up and went back to windows. >high-end robot vacuums use LIDAR Really? That's interesting. I know the scanners from work, but they're expensive as fuck. I figured LIDAR and RGBD cameras were outside my reach and I had to settle for regular rbg, maybe stereo. With slam3r that might even be possible, but I read that the calculated camera position isn't very accurate. Eh, I'll figure it out along the way. I think the pathfinding won't be a problem. I made a game once with voxel based pathfinding. Wasn't so great because the map was huge and I wanted hundreds of enemies without collisions. But a small map with static obstacles should be manageable. The bigger problem will be the animations. I tired some Quest 3 apps. Where you can decorate your home with virtual furniture lol. Making some crazy animation system for furniture or advanced IK would be badass. >I like your idea. You're clearly very knowledgeable and talented. Thanks :3 I like my crazy projects!

Updated the Galatana description to tell how she functions to make it more clear to new users https://greertech.neocities.org/galatana @rick , any updates?

Contrived (but compelling nonetheless) promo video of a Gatebox AI ripoff clone. https://youtu.be/QtAIiJ1wIIc Dipal D1 looks very interesting at this stage. I wish them good success for now!

>>40250 Much-better look into the product from the consumer's viewpoint. https://www.youtube.com/watch?v=xZiwjLF7zTU

>>40250 I recently learned Godot is able to run on raspberry pi 4 and up. Since it is an engine capable of 3d graphics it would be possible to make an animated avatar in godot and have it interact with Ollama via a .txt file. Haven't tried godot yet so I'll let ya know how it works out. Would probably be easier than trying to hard-code everything in python, if there isn't an outright ollama plugin lol

>>40281 >Haven't tried godot yet so I'll let ya know how it works out. Exciting! Godspeed, Mechnomancer.

>>38612 >>40250 These seems overpriced for what it offers, and having a subscription fee makes me assume it's going to be yet another overpriced AI chatbot service that you have to worry about disappearing. I would think it wouldn't be too hard to assemble something like this out of existing technology. It feels like there's a ton of small projects that try to offer something like this but are all split up and don't work with each other. For the software, SillyTavern can likely handle all of this. It's self-hosted, has Live2D and VRM support, has TTS and voice recognition support, and you choose the LLM backend. If you paired that with outputting the model display to a smartphone/small monitor in a pepper's ghost box, you'd have a rudimentary version of this. I think the only problem would be getting ST to output the model on a separate display. https://www.youtube.com/watch?v=WRTWyPXYotE I'm gonna look around to see if there's any good 3D printable pepper's ghost projects, because the software seems do-able with existing projects. (Hope I'm not coming across as too much of an ideas-guy. I check this place out a couple times a year and I've recently gotten into AI roleplaying with ST. I'm looking into a dedicated visual display for a hologram waifu, and the recent rise in LLM capability makes all this seem much more viable)

>>41756 >(Hope I'm not coming across as too much of an ideas-guy. I check this place out a couple times a year and I've recently gotten into AI roleplaying with ST. I'm looking into a dedicated visual display for a hologram waifu, and the recent rise in LLM capability makes all this seem much more viable) What!? No, you're fine Anon! This idea of yours is great. I think if you look around the board (even here ITT) you'll find some good ideas about Pepper's -like systems. I personally consider this one of the more-achievable DIY-style projects for Anons to attempt, so I was really in favor of OP starting this thread. Please keep us all up to date with your research progress, Anon! Cheers. :^)

>>41756 >building pepper's ghost with a smartphone Pepper's ghost is very easy: see the pic, anon. You can also scale it up for larger screens: it is simply a transparent piece of plastic placed above the screen at a 45 degree angle. Happy building!

>>41766 Great example of KISS! We here will need to be creative like this to keep costs very low. That will be important to spreading robowaifu tech far & wide around the globe! Cheers, Anon. :^)