This is so sad, people work so hard to make Vtuber models and takes weeks even months. It’s really stupid how people, even apps in this case are using Ai.. why, when they Ai models even looks so bad help. Please make your own, there is nothing wrong with buying a model from someone but this is really wrong especially selling Ai art… GET AI AWAY FROM ME.
i hate the legalese jargon they use in their blog post too. they say the *illustrations* were not made with generative ai, but what about the movement mechanics? the only way i can see having one (1) singular illustration have such fluid movements without being rigged in live2d is to use generative ai to fill in the blanks
Agree, the way they speak specifying everything feels very much like they just don’t want to be sued or lose business (especially when you can literally just look up every law in existence on google, they’re very cautious of everyone ) more than they actually want to help resolve the issue for the people who they gave the models to Also cool paruko pfp :3
basically, and i mean reaaaalllyyy basically, their app has like a "live photo filter" that can be used to make 2d illustrations move and blink. it's obviously going to have some use of AI/generative AI to make that happen as the program will have to "fill in the blanks" of the illustration going from the original illustration to another pose/having movement like the blinking and mouth movement.
this just wha happens when u feed a community to believe anyone can be an idol as long as they can afford a pretty anime model. dont get me wrong a lot of vtubers are just artists who want to stay anonymous, but u can find A LOT of vtweeters/vtubers who just want the fame and treatment as an idol whilst being good at nuffin in particular. besides stirring drama.
Did you ever address the video @Rev says desu made about you spreading misinformation about the loli vtuber group? I just wanted to come here and see as I just happened upon his video about you and I don’t see a specific video responding to his.
It's a half truth. They may not have used generated models, but to resort to using ages old flash style "rigging" (which, if I had to guess is what they're using) cuts out riggers and obstensively disrespects that entire group of artists. To go more into detail, what it sounds like they're doing is basically mapping triggers to automated transformations. Take a png, put some sort sensory map over it where the eyes match up with the eyes, mouth with the mouth, etc. Then just like how a rigged model may react to tracking data to move individual parts, it will distort the png to make it look like the same effect. For example, blinking takes whatever the color in the top third eye is, assumes it's the eyelid color, and increasingly colors the eye region as the eyes close. Flash animation used to do something similar, although with key differences. For one you'd never use a png - pixels in a raster images have both an x and y coordinate and things get very quickly weird if those are ever disturbed in all but a handful of ways. Instead, you used vector - whcih, through the power of math, kept shapes nicely and could translate them across multiple transformations. You'd still also cut up the model into different parts, to not only make sure jointed segments stayed together, but to more concisely ensure movement looked believable. It's all to say that this doesn't have to involve machine learning or anything more advanced than flash animation had, and it would still make no difference. In a vtubeing space where a large portion of the art comes from the meticulous cutting and rigging of intricate models, cutting corners is basically disrespect.
i use a model from vroid, (drew the colour of the clothes myself and how the eyes and the lips and the eyebags are, but i already feel bad that i still cant draw an fully own model because i cant draw that on phone with finger(tried so often but its way to hard) so the first thing ill do if i ever get money from streaming to buy something to draw my own model, thats why i personally would, even if i get paid, never join such stuff, because i want to be myself and not something i dont feel iam smh
u can prolly use a base whether its free or paid (from actual artists) to reference a full body and use that to roughly design ur model for now. a good thing about designing is that u dont have to draw anything perfectly if its not the final product. i suggest this just so ure not waiting so long and feeling stuck wit no process. i am also going through the process of investing on things to start creating content. this is just an idea as i relate with you and im appreciative that people like u exist, people who DONT run to ai because theyre too afraid to try. its also nice to hear about the realness behind wanting to become a vtuber and the struggles of prepping for it financially and artisticly. anyway, from one to another i wish u luck and want to send positivity towards your way. u can do this and no matter what, never stop practicing in little ways for your big debut.
This service will fail, the appeal of vtubers are the models. Whatever they use to rig the models looks so uncanny its unappealing, with the still body and twitching face. They failed at step one.
so it's like that live photo filter? sounds like they're taking a picture drawn by an actual artist and then the program takes that and tries to animate that? technically that would be an ai program but that wouldn't be ai art. only halfway through the video so idk if there will be further explanation but that sounds like a good program for people that couldn't afford a fully rigged custom model. if you could just take your own art, or commissioned art, and use it in the app with that program it would open a lot of doors. though i imagine riggers would of course be pissed if people stopped paying them to rig.
They're bascially laying a heat map over the image and the program is warping what it is being told are the eyes, mouth, etc. It's not AI, but that doesn't really matter. Vtubeing gets the look that it does by riggers segmenting a model into layers (easily hundreds of layers depending on the model) which can then be individually affected by the mocap. What these dipshits are doing is tryign to cut the riggers out of the equation.
I agree. This is nice alternative for people who already have an OC and have made art for it but don't know how to rig or have not found an artist and/or rigger within their budget to commision an actual Live2D model. The obvious quality issue is obviously going to be noticed. We've been using AI face recognition in our daily lives a lot anyways from Face ID to beauty filters and other fun filters used in social media apps. Not everything is bad with AI. AI was originally meant as a tool to help humans anyways, not take over our jobs. Like this app will not be taking over any rigger's job because of the obvious quality difference. Except for actual AI ART that's actually stealing from artists as their training data since art theft has been a problem for millennia.
This is NOT at ALL the main problem with anilive. They've treated their talents horribly! please please look for karmunism's video on their experience. They were one of the talents employed by anilive and that situation that she and the others went through is what should be highlighted here!
Hi! I'm an ex anilive talent who made a 45 minute deep dive on the company. It has been blowing up on youtube and Twitter. I'm deeply disappointed in the lack of research you display in this video. You leave out so much context, and due to that spread a lot of misinformation. All you did was read people's tweets and a couple other publicly available resources, then make some speculative comments. This video feels almost insulting considering how much work some of us talents put in to share the actual truth. Even a little bit of research could have been done to clarify some of the things in your video, especially on the AI part. You were super misleading about how this was addressed. The art itself is human made while the in house rigging program animates the static png. This animation process is where the artifaxting and other issues come in. You can see the difference between the animated art and the original art. It's just little things like that which are super frustrating. No context is given whatsoever. This isn't the first time you've made a video without researching it, and it landed you in hot water back then too. Do better.
Yet that animation bit is where the allegations hold the most water. There is missing information (I would still suggest watching both vods) but the "in house rigging program" is doing the work that riggers would have done, and done better. That combined with the job posting and the fact they wouldn't even need Machine Learning to create an incredibly trite animation function that just deforms a singular PNG is enough to make all the assurance in the world null and void. It's still scummy and, considering the other things you pointed out in your video, suspicious to boot.
Yeah, I noticed the models don't have the mistakes the IA normally do, in general, they seen concise and the continuity in the drawing looks good. What feels to be the problem is the rush work and cheap rigging, due to the platform limitations or them being free (a bit scummy). Also, if the company have work with artists, I understand not wanting to tell who they are (for contractural restrigements or to avoid the IA witch hunt to affect them)
This is the main main reason why I skipped this video. Not even a trace of what you guys went through and what the real problem with anilive is. This is like strike two in the same week for kat for me. She's so focused on sensationalism that she's not doing research :(
Soooo btw, an ex liver made full video with so much more information. I recommend checking that out, has so much more in depth research.… arguably better context shown there
Kat I really enjoy your videos and takes, but I have to be honest your videos about AI seem very ingenuine knowing you promote Displate, a company who constantly uses AI for their products. On twitter you literally posed with a poster which had the ugliest AI cat i've ever seen. You can't come on here and talk about the harm AI does to real artists, taking away business and money, (even though that's all true) but then go on to profit from AI yourself. Please think about artists like yourself, and make your stance firm or just don't speak about the topic at all perhaps.
Right! I was waiting for someone to say it. It’s disgusting that she’s sponsored and profiting off of AI while bashing other VTubers and artists for using AI.
As a vtuber artist, this physically hurts. You can see they just put a layer over the ai and "animated it" to cover and not cover it. Obviously cutting corners to not pay artists.
8:21 okay, one ex-AniLive jumpstart member confirmed that they did choose but AniLive only gave them a few choices and was very strict on the changes like you could only change some things like little details like colour of clothing or little details like add horns
9:50 btw nexus which is a agency which is linked to this app but is not the same entity responded on. We do not use AI and we do not support it before AniLive made this post, even though Nexus wasn’t under fire for the AI
In the last two years, and because of AI the art community started to offer the same entertainment value as the flat earthers, the woke mob, and the letter people. Hollywood is nowhere even close at this point. I love watching the entire art community go down in flames. You people are fighting among yourselves like there is no tomorrow with even the slightest hint of AI being involved, it's hilarious. But I don't think content creators hate the extra views, so keep it up, keep bringing the entertainment in such a well-made format.
i'd recommend viewing karmunism's video on this topic! she's a former anilive jumpstart volume 2 talent and she clarified that the illustrations aren't ai, just a rush job (hence why there are some dots that didn't get erased, haphazardly drawn patterns, inconsistent line thickness, lack of fingernails, eye whites not being filled in, etc.) but the rig is absolutely done with ai (which is the "feature" anilive mentioned in their statement about "allowing pngtubers to become vtubers with just an illustration") and much of the artifacting, loss of detail, stiffness, strange movements that wouldn't be present on an actual live2d model, etc. are a result of the model not being split into parts the way a live2d model would and essentially just being a drawing made by a human but made to move and blink with ai templates. karma also details many other issues with the company, such as onboarding too many talents while not providing adequate support to the talents they already have or hardly promoting them at all (not even retweeting their tweets until like 4 days had passed since the thing the tweet was talking about), taking forever to reply to messages/emails and to release a statement about the ai thing while anilive jumpstart talents were already getting flack for it despite it not being their fault, and generally being very unprofessional even outside of the whole ai thing. like karma got an email that just said "i sent you a friend request" with no signature or anything after letting anilive know four times in three different ways that she wasn't able to add the staff member she was told to add on discord and being completely ghosted for days leading up to her interview. but i'd also like to point out that the other option of uploading the live2d model you already have directly into the app as opposed to using vtube studio and essentially giving the company full access to a model you paid potentially thousands of dollars for is also shady as hell and highkey sounds like a scam, even ironmouse said as much five months ago
They had messaged me about their vtuber program thing on twitter. And invited me to it. Said they'd provide me with everything i needed and they would pay me to stream for them. I am now honestly kind of happy i didn't go through with it. Even though i was really considering it. I did research on everything too with a friend. Nothing seemed wrong everything seemed too good to be true. I did see some models being a bit stiff, no ai or anything. But i now am noticing all these people having issues with the second wave of vtubers. It being ai, stiff and generally not really polished. They allow u to use Ur own model. which is probably the ones u were shown on their profile. the no credits to artists. in a way is just kind of weird. the characters looking similar to others always CAN happen but it happening over and over was very suspicious. IM so so glad i decided against it ;; even when the offer was very very interesting
I'mma be honest, if a company "offered" me a vtuber model to stream on their platform and it's rigged like that, I'm walking. From the looks of it, IF it's not AI, it's either incredibly rushed or they don't know what they're doing. Upon receiving all these inquiries, AniLive should have realised something as wrong.
As someone who’s a indie vtuber who streams on twitch, also on anilive, this isn’t always what these models looked like. The first few vtubers they handed out looked fine. They’re just pushing out SO MANY vtubers who only stream on this app, they’re just doing SO MUCH and need to slow down. I would also like to say for you to actually be able to stream on the app, you need to send your model to them in a smaller file, apply to even be on the app as a streamer, and more. It’s such a long process to even stream on this app. Not anyone can stream on this app. So, long story short they’re forcing out so many vtubers so fast and should slow down and focus on their existing talents that are already on the app. NOT ALL THE VTUBERS ON THIS APP ARE LIKE THIS.
4:10 you may not know this, but there was some models which are just rigged images like you can find a few in the workshop of VTube studio and also most of the time horns are not supposed to be rigged since you know horns are solid like you don’t get that much movement with horns actually you get no movement the only movement of the horns would be head movement sense unless it’s something like sort of tentacle horns they are not supposed to move because in nature horns are supposed to be solid and be able to like do injury to things like any frats so I would think unless a artist is going for the tentacle Horn they wouldn’t move doesn’t matter if it’s a rig of a model or a animation
They aren't even models it's horrible it's not ai art but they take an artist png they made for anilive and then a program distorted the art to make a png so not a live2d model move "like" a model. So they are using an ai program to rig a png not a cut live2d model art a flat png. I feel so bad for the artist that their art is now being called ai and they aren't even getting credited. Apparently there is no contract for them so they are not obligated to stay or keep quiet from what I have seen from ex anilivers. They are cheap and horrible. They would have been better off giving them a character design and made it on a premade customizable model. An ex liver that left before debut showed their art before and after it was put in the ai program and it looks real and normal before and fucked up and ai after. It's really sad I already didn't trust this app when it came out.
@@RobertStoll Yeah it seems something like that then but done by an ai program definitely looks bad because it's not human made. I think it would be more acceptable if they were more transparent because they are just scamming a bunch of people who signed up at this point.
Heyaa!! Ex-anilive jumpstart here! I have used AniLive since Jumpstart 1. I had quit being on of their jumpstarts and deleted the app. I had been going live for months without knowing about this. When my sister also applied for AniLive, she was given a VTuber model but guess what? It looked exactly like mine, I mean not COMPLETELY but REALLY similar, this made me so upset and frustrated because I had work for a WHOLE year on my vtuber model so no way i'd let them do that! I went live and talked about it to a few people(a few ppl idk since I barely got any attention or viewers) And at this point of the time, Im siding with the new jumpstarts who got AI-looking generated model. This is literally unfair and I feel bad for them, I had created a video about this but im currently re-doing it because my voice was cracky and hoarse. I feel bad for the VTuber artist who worked/who was hired for/by Anilive, they really should compensate them because why would they take advantage(I don't know how else to word this so..) of them? That's literally just cruel, I can relate to the artist that she was naive because I was too.
Reminds me of the website/free Steam game that let you "make" 3d vtubers and they were all stolen paid models from Booth and Gumroad. On their official website they were even reselling the models and acting like they owned them
5:50 - i would be surprised, if the company would even win a single % of a lawsuit at this point. At least in the EU contracts can be voided in specific cases and the company intending to ruin the VTubers reputation is quite a strong argument for that.
If anything, they could have commissioned artists, but could be using AI for editing/transitions. But the fact that they won't name the artists tells volumes.
You know what they look like? Those tiktok ai filters, like the ones that will make you 'look like a disney princess' and it just slaps it over your face while you talk and move
Honestly weather its ai or not its still really crappy as someone who is an artist but doesn't make models or riggs its still pretty bad ngl. again even if its not ai its still scummy to give people who are most likely newer to vtubing a model that just has that obvious of issues i get that its free but still. edit: finished the whole vid and point still stands however i also wanna add its just decitful to give someone a quickly made model that doesnt even have the same functions of a normal pngtuber
This absolutely could have been a case of AI animation on real illustrations. AI animation is really janky still and absolutely would add those AI traits to the art because it makes use of generative AI. I don't think that's any better than using AI art honestly, though. And we haven't seen any still illustrations without the artifacts yet...and they were hiring gen AI engineers
They invited me to the jumpstart program and I thought it was weird they required your model's files if you already had one, now that I know they use AI I can tell why. They were trying to train their shitty AI with it LMAO. Big red flags so thankfully I didn't join/give them my files
The Live2D rigger in me is screaming watching this. A human would NEVER look at this and say "eh good enough". There's a reason Live2D models are expensive and this is a great example of what it looks like when you try to cut costs. Also I remember drama with this app in the Live2D community some time ago. They asked for the mco3 files (editable project files in Live2D) from every VTuber who signed up which is a massive concern because it means they can temper with the rig and use it as a template. Most riggers don't give out project files for that reason, we don't want our hard work getting messed with or copied.
oh, how funny, this was the same company that had a controversy about how their app operated idk if it still works that way now, but basically, you had to UPLOAD your model files to their website, and the way they do the tracking of your movements is by sending the tracking data from YOUR phone to THEIR servers this differs highly from how vtubers usually do it because these two things are done all on your computer, so this raised concerns about them maybe using the model files for AI training, and also privacy concerns from people thinking they uploaded your camera image to their servers which also meant that they don't livestream VIDEO output to viewers devices... they apparently stream the model and tracking data directly! which means that when someone opens a vtuber stream, their model gets DOWNLOADED to their device, and someone could theoretically steal it if they know how!! isn't that fun?? /sarc edit: fixed typo
Here's the thing: They're advertising themselves as a VTuber app, not a PNGTuber one, or some general streaming service. Sure, they mentioned getting a free ANIME CHARACTER, not a VTuber model, but it doesn't make it any better, because they specifically emphasize on VTubing. And while we're at it, the animations and rigging are HORRIBLE. Not only are the models probably a mix of AI and very poor adjustments, but the movements themselves don't make any sense whatsoever. It looks so bad and I'm so sorry for the people that received THIS as their models to debut with. I'm so sure it's not what they were promised, or what they believed they would receive, and it can very well be considered false advertisement or deception Now, how can they be so sure that the "illustrations" aren't AI? They're either lying, or they were scammed into believing that the images are legit. They don't have any means of knowing if the people they supposedly commissioned didn't or did use AI. All of this is so sketchy I mean, that's what happens when a company claims to give you something for free. If they're giving away "free models", then the quality will be representative of that, and as we can see, it is very much bad
this video was just made to churn out content, wasn't it? zero research, the same three false claims over and over again, and no regard for what the actual streamers were saying. don't leave this up. it's blatant misinformation, really mean towards the actual artists that drew those avatars, and every single vtuber affected by the app is in these comments telling you that this isn't the main issue about this app. also please just stop "analysing" ai stuff. you keep fumbling more and more with each video, where you call out stuff that isn't even an issue with ai anymore and confuse it with human mistakes or, in this case, with a different way to animate that doesn't use rigs. you also don't see to have an issue taking displate money, when they are so riddled with ai images - so ig you only care if you can make money off of it.
Let's leave out the "rigging" part of that feature. The art is still clearly AI. Not only that but there are amazing png model that look incredible and way better than that slop. Like they could've commissioned an actual PNG artist for a good price.
its like in between png tubing nd vtubing fully wit the rigging, like someone jus day 1 started rigging their models wit improper seperations for the art 💀 its choppy as u can pause every single frame and its so stiff when it sways. these vtubers simply wanted an opportunity but was given this nonsense and not what they potentially deserve. i cant even say they should keep their design of their models that was given, because it was most likely stolen through the ai too.
Most of model is cheap, but its not AI (yet), it has the same beginner mistakes I've made while drawing. The closest one to AI is the purple dress girl with the dress being held like a royalty, but its not yet conclusive. You can call it cheap or rushed or lacked quality control, but not yet AI. You should have put the Generative AI Engineer hiring part from the LinkedIn on the front because that's the meat of the issue.
18:15 is this a fully rigged model? What is it the half vtuber Half pngtuber model if it’s fully I don’t believe it’s made by AI because AI is too stupid to make a fully riggable model but if it’s a rigged png model I guess that would be possible
It doesn't look like genAI art, more like that kind of dancing filters where you can stick your picture and they make your photo dance. I think they're just scrimping on money tbh. No one is just gonna give out free live2d models. Also if you're a company/org the monthly fee you need to pay to Live2d/Cubism is really high loll. Using this kind of cheap filter is how they're giving 'models' out for free. As for the artist/crediting, nijisanji allows the artists to claim 'mama'-ship but A LOT of their old stock (they have a huge stash of models that they just pick out to use and distribute) do not have mamas and are under nijisanji's ownership. They just pay artists to draw designs and rig it in-house back then. In Japan they rarely credit the artists, esp in those 2d card gacha games, unless you're a really famous/popular artist, then they'll plaster your name everywhere and call it a collab lol. A lot of your fav JP artists are doing contract work outside (ever notice they disappear for months then come back and start drawing your couple again?) but they are never credited or have NDAs. I am in no way defending Anilive, just discussing how the real art industry works outside of twitter vtubers and artists lol
man, anilive sounds great but its such a downer if they give models like this, like if you dont wanna go through the process of making a legit model due to lack of money and time or just being lazy, offer a PROPER non-AI PNG-tuber model to start them with, that's still technically a Vtuber model AND its much easier and quicker to produce I feel so bad for the person who got this model and rig as their prize, at least it was free i guess...but this is just a scam
Say you want to animate an eye blinking. For the best effect, you could draw an open eye and a closed eyelid. Then layer the closed eyelid on top of the open eye whenever the eye was "closed." This is how most vtubers nowadays are rigged; it's like an intricate digital puppet with layered parts that can be effected by the face tracking. If you wanted to cut a corner, however, you could just take that open eye and draw and note the area around the eye, along with the color of the eyelid. When the eye closes, color in that area with the color of the eyelid, and when it is open don't color it in. That's what these guys are basically doing. Flash animators used to do the same, only it looked really stilted unless you put effort in and, well, cut your model up into multiple pieces.
If it was real art, not ai. Why wouldn't they say names? Understand artist want to maintain anonymous but kf people are saying your art is ai i imagine they would come forward
I put this in another comment's reply but this app (not their Jumpstart program) is a nice alternative for people who already have an OC and have made art for it but don't know how to rig or have not found an artist and/or rigger within their budget to commision an actual Live2D model. The obvious quality issue with an actual just normal PNGtuber model(? is that the right word? im not sure) with proper blink art and/or expressions is obviously going to be noticed. But maybe the artist is charging an ungodly amount just to make another art piece that's basically the same just with closed eyes and/or different mouth shape. We've already been using AI face recognition in our daily lives a lot anyways: from Face ID to beauty filters and other fun filters used in social media apps. Not everything is bad with AI. AI was originally meant as a tool to help humans anyways, not take over our jobs. Like this app will not be taking over any rigger's job because of the obvious quality difference. Except for actual AI ART that's been generated from an AI Art Program since that's actually stealing from artists as their training data. It's unfortunate art theft has been a problem for millennia.
I really hate how ai bros are trying to convince artists they need to intergrate ai into their art or the “will be left behind” I remember a tracer whos art i was investigating because they traced before you could look at it and see ai traits that they didn’t correct. (Unfortunately they’re a massive comment deleter and delete comments that don’t agree)
Vtuber model ≠ Pngtuber model. Looks like that App is for Pngtuber models with limited rig on eyes and mouth (Made with neural networks maybe?) Also first models looks like hand made not AI. But you need to take in consideration that maybe are just concept art for the model that was just rigged for pngtubing. Not everybody have 700USD for a full rigged Vtuber model and some just could use this for curiosity and not a "necessity".
Are there companies like this out there that aren't so.. controversial? I only hear about these company ideas that I love and would want to support, on drama channels.. And they only get on drama channels once they've done something wrong🥲
This is so sad, people work so hard to make Vtuber models and takes weeks even months. It’s really stupid how people, even apps in this case are using Ai.. why, when they Ai models even looks so bad help. Please make your own, there is nothing wrong with buying a model from someone but this is really wrong especially selling Ai art… GET AI AWAY FROM ME.
Too bad they dont care about quality.
They care only how much time and money they can save
Ai bro is cheap
@@grildragoI extremely agree. To add on, VTuber models are like thousands of $ 😭
I literally downloaded this app and hadn't even realized any issue, thank you Kat for saying something I was about to fall in that trap😂
i hate the legalese jargon they use in their blog post too. they say the *illustrations* were not made with generative ai, but what about the movement mechanics? the only way i can see having one (1) singular illustration have such fluid movements without being rigged in live2d is to use generative ai to fill in the blanks
Agree, the way they speak specifying everything feels very much like they just don’t want to be sued or lose business (especially when you can literally just look up every law in existence on google, they’re very cautious of everyone ) more than they actually want to help resolve the issue for the people who they gave the models to
Also cool paruko pfp :3
basically, and i mean reaaaalllyyy basically, their app has like a "live photo filter" that can be used to make 2d illustrations move and blink. it's obviously going to have some use of AI/generative AI to make that happen as the program will have to "fill in the blanks" of the illustration going from the original illustration to another pose/having movement like the blinking and mouth movement.
Me eating up all the Vtuber drama that seems to appear every week or so now
So so real
this just wha happens when u feed a community to believe anyone can be an idol as long as they can afford a pretty anime model.
dont get me wrong a lot of vtubers are just artists who want to stay anonymous, but u can find A LOT of vtweeters/vtubers who just want the fame and treatment as an idol whilst being good at nuffin in particular. besides stirring drama.
oops ignore the fact i uploaded 2.5 hours late i forgot to make a thumbnail last night LOL
Ermm rovio allegedly used AI on their newest game
its fine Kat, ur a human too, i dont think ppl would go angry af just because of that, take breaks too if u need them ^^
It’s okay
it’s okay
Did you ever address the video @Rev says desu made about you spreading misinformation about the loli vtuber group? I just wanted to come here and see as I just happened upon his video about you and I don’t see a specific video responding to his.
It's a half truth. They may not have used generated models, but to resort to using ages old flash style "rigging" (which, if I had to guess is what they're using) cuts out riggers and obstensively disrespects that entire group of artists.
To go more into detail, what it sounds like they're doing is basically mapping triggers to automated transformations. Take a png, put some sort sensory map over it where the eyes match up with the eyes, mouth with the mouth, etc. Then just like how a rigged model may react to tracking data to move individual parts, it will distort the png to make it look like the same effect. For example, blinking takes whatever the color in the top third eye is, assumes it's the eyelid color, and increasingly colors the eye region as the eyes close.
Flash animation used to do something similar, although with key differences. For one you'd never use a png - pixels in a raster images have both an x and y coordinate and things get very quickly weird if those are ever disturbed in all but a handful of ways. Instead, you used vector - whcih, through the power of math, kept shapes nicely and could translate them across multiple transformations. You'd still also cut up the model into different parts, to not only make sure jointed segments stayed together, but to more concisely ensure movement looked believable.
It's all to say that this doesn't have to involve machine learning or anything more advanced than flash animation had, and it would still make no difference. In a vtubeing space where a large portion of the art comes from the meticulous cutting and rigging of intricate models, cutting corners is basically disrespect.
i use a model from vroid, (drew the colour of the clothes myself and how the eyes and the lips and the eyebags are, but i already feel bad that i still cant draw an fully own model because i cant draw that on phone with finger(tried so often but its way to hard) so the first thing ill do if i ever get money from streaming to buy something to draw my own model, thats why i personally would, even if i get paid, never join such stuff, because i want to be myself and not something i dont feel iam smh
u can prolly use a base whether its free or paid (from actual artists) to reference a full body and use that to roughly design ur model for now. a good thing about designing is that u dont have to draw anything perfectly if its not the final product. i suggest this just so ure not waiting so long and feeling stuck wit no process.
i am also going through the process of investing on things to start creating content. this is just an idea as i relate with you and im appreciative that people like u exist, people who DONT run to ai because theyre too afraid to try. its also nice to hear about the realness behind wanting to become a vtuber and the struggles of prepping for it financially and artisticly. anyway, from one to another i wish u luck and want to send positivity towards your way. u can do this and no matter what, never stop practicing in little ways for your big debut.
This service will fail, the appeal of vtubers are the models. Whatever they use to rig the models looks so uncanny its unappealing, with the still body and twitching face. They failed at step one.
so it's like that live photo filter? sounds like they're taking a picture drawn by an actual artist and then the program takes that and tries to animate that? technically that would be an ai program but that wouldn't be ai art. only halfway through the video so idk if there will be further explanation but that sounds like a good program for people that couldn't afford a fully rigged custom model. if you could just take your own art, or commissioned art, and use it in the app with that program it would open a lot of doors. though i imagine riggers would of course be pissed if people stopped paying them to rig.
They're bascially laying a heat map over the image and the program is warping what it is being told are the eyes, mouth, etc.
It's not AI, but that doesn't really matter. Vtubeing gets the look that it does by riggers segmenting a model into layers (easily hundreds of layers depending on the model) which can then be individually affected by the mocap. What these dipshits are doing is tryign to cut the riggers out of the equation.
I agree. This is nice alternative for people who already have an OC and have made art for it but don't know how to rig or have not found an artist and/or rigger within their budget to commision an actual Live2D model. The obvious quality issue is obviously going to be noticed. We've been using AI face recognition in our daily lives a lot anyways from Face ID to beauty filters and other fun filters used in social media apps. Not everything is bad with AI. AI was originally meant as a tool to help humans anyways, not take over our jobs. Like this app will not be taking over any rigger's job because of the obvious quality difference. Except for actual AI ART that's actually stealing from artists as their training data since art theft has been a problem for millennia.
This is NOT at ALL the main problem with anilive. They've treated their talents horribly!
please please look for karmunism's video on their experience. They were one of the talents employed by anilive and that situation that she and the others went through is what should be highlighted here!
❤
Hi! I'm an ex anilive talent who made a 45 minute deep dive on the company. It has been blowing up on youtube and Twitter.
I'm deeply disappointed in the lack of research you display in this video. You leave out so much context, and due to that spread a lot of misinformation. All you did was read people's tweets and a couple other publicly available resources, then make some speculative comments.
This video feels almost insulting considering how much work some of us talents put in to share the actual truth. Even a little bit of research could have been done to clarify some of the things in your video, especially on the AI part. You were super misleading about how this was addressed.
The art itself is human made while the in house rigging program animates the static png. This animation process is where the artifaxting and other issues come in. You can see the difference between the animated art and the original art.
It's just little things like that which are super frustrating. No context is given whatsoever. This isn't the first time you've made a video without researching it, and it landed you in hot water back then too. Do better.
Yet that animation bit is where the allegations hold the most water. There is missing information (I would still suggest watching both vods) but the "in house rigging program" is doing the work that riggers would have done, and done better.
That combined with the job posting and the fact they wouldn't even need Machine Learning to create an incredibly trite animation function that just deforms a singular PNG is enough to make all the assurance in the world null and void. It's still scummy and, considering the other things you pointed out in your video, suspicious to boot.
Yeah, I noticed the models don't have the mistakes the IA normally do, in general, they seen concise and the continuity in the drawing looks good. What feels to be the problem is the rush work and cheap rigging, due to the platform limitations or them being free (a bit scummy).
Also, if the company have work with artists, I understand not wanting to tell who they are (for contractural restrigements or to avoid the IA witch hunt to affect them)
This is the main main reason why I skipped this video. Not even a trace of what you guys went through and what the real problem with anilive is. This is like strike two in the same week for kat for me. She's so focused on sensationalism that she's not doing research :(
This comment needs more likes.
Boosting this comment, I watched your video and it was really informative! I hope this video gets corrected :(
Soooo btw, an ex liver made full video with so much more information. I recommend checking that out, has so much more in depth research.… arguably better context shown there
❤
Kat I really enjoy your videos and takes, but I have to be honest your videos about AI seem very ingenuine knowing you promote Displate, a company who constantly uses AI for their products. On twitter you literally posed with a poster which had the ugliest AI cat i've ever seen. You can't come on here and talk about the harm AI does to real artists, taking away business and money, (even though that's all true) but then go on to profit from AI yourself. Please think about artists like yourself, and make your stance firm or just don't speak about the topic at all perhaps.
Right! I was waiting for someone to say it. It’s disgusting that she’s sponsored and profiting off of AI while bashing other VTubers and artists for using AI.
16:56 Her name is Navia and everyone loves to call her Tailor Swift
15:25 Oh and that is Albedo from Overlord
As a vtuber artist, this physically hurts. You can see they just put a layer over the ai and "animated it" to cover and not cover it. Obviously cutting corners to not pay artists.
8:21 okay, one ex-AniLive jumpstart member confirmed that they did choose but AniLive only gave them a few choices and was very strict on the changes like you could only change some things like little details like colour of clothing or little details like add horns
❤
9:50 btw nexus which is a agency which is linked to this app but is not the same entity responded on. We do not use AI and we do not support it before AniLive made this post, even though Nexus wasn’t under fire for the AI
In the last two years, and because of AI the art community started to offer the same entertainment value as the flat earthers, the woke mob, and the letter people. Hollywood is nowhere even close at this point.
I love watching the entire art community go down in flames.
You people are fighting among yourselves like there is no tomorrow with even the slightest hint of AI being involved, it's hilarious. But I don't think content creators hate the extra views, so keep it up, keep bringing the entertainment in such a well-made format.
i'd recommend viewing karmunism's video on this topic! she's a former anilive jumpstart volume 2 talent and she clarified that the illustrations aren't ai, just a rush job (hence why there are some dots that didn't get erased, haphazardly drawn patterns, inconsistent line thickness, lack of fingernails, eye whites not being filled in, etc.) but the rig is absolutely done with ai (which is the "feature" anilive mentioned in their statement about "allowing pngtubers to become vtubers with just an illustration") and much of the artifacting, loss of detail, stiffness, strange movements that wouldn't be present on an actual live2d model, etc. are a result of the model not being split into parts the way a live2d model would and essentially just being a drawing made by a human but made to move and blink with ai templates. karma also details many other issues with the company, such as onboarding too many talents while not providing adequate support to the talents they already have or hardly promoting them at all (not even retweeting their tweets until like 4 days had passed since the thing the tweet was talking about), taking forever to reply to messages/emails and to release a statement about the ai thing while anilive jumpstart talents were already getting flack for it despite it not being their fault, and generally being very unprofessional even outside of the whole ai thing. like karma got an email that just said "i sent you a friend request" with no signature or anything after letting anilive know four times in three different ways that she wasn't able to add the staff member she was told to add on discord and being completely ghosted for days leading up to her interview. but i'd also like to point out that the other option of uploading the live2d model you already have directly into the app as opposed to using vtube studio and essentially giving the company full access to a model you paid potentially thousands of dollars for is also shady as hell and highkey sounds like a scam, even ironmouse said as much five months ago
They had messaged me about their vtuber program thing on twitter. And invited me to it. Said they'd provide me with everything i needed and they would pay me to stream for them.
I am now honestly kind of happy i didn't go through with it. Even though i was really considering it. I did research on everything too with a friend. Nothing seemed wrong everything seemed too good to be true. I did see some models being a bit stiff, no ai or anything. But i now am noticing all these people having issues with the second wave of vtubers. It being ai, stiff and generally not really polished. They allow u to use Ur own model. which is probably the ones u were shown on their profile. the no credits to artists. in a way is just kind of weird. the characters looking similar to others always CAN happen but it happening over and over was very suspicious. IM so so glad i decided against it ;; even when the offer was very very interesting
I'mma be honest, if a company "offered" me a vtuber model to stream on their platform and it's rigged like that, I'm walking.
From the looks of it, IF it's not AI, it's either incredibly rushed or they don't know what they're doing.
Upon receiving all these inquiries, AniLive should have realised something as wrong.
As someone who’s a indie vtuber who streams on twitch, also on anilive, this isn’t always what these models looked like. The first few vtubers they handed out looked fine. They’re just pushing out SO MANY vtubers who only stream on this app, they’re just doing SO MUCH and need to slow down.
I would also like to say for you to actually be able to stream on the app, you need to send your model to them in a smaller file, apply to even be on the app as a streamer, and more. It’s such a long process to even stream on this app. Not anyone can stream on this app.
So, long story short they’re forcing out so many vtubers so fast and should slow down and focus on their existing talents that are already on the app. NOT ALL THE VTUBERS ON THIS APP ARE LIKE THIS.
this is so messed up 😨😨
4:10 you may not know this, but there was some models which are just rigged images like you can find a few in the workshop of VTube studio and also most of the time horns are not supposed to be rigged since you know horns are solid like you don’t get that much movement with horns actually you get no movement the only movement of the horns would be head movement sense unless it’s something like sort of tentacle horns they are not supposed to move because in nature horns are supposed to be solid and be able to like do injury to things like any frats so I would think unless a artist is going for the tentacle Horn they wouldn’t move doesn’t matter if it’s a rig of a model or a animation
They aren't even models it's horrible it's not ai art but they take an artist png they made for anilive and then a program distorted the art to make a png so not a live2d model move "like" a model. So they are using an ai program to rig a png not a cut live2d model art a flat png. I feel so bad for the artist that their art is now being called ai and they aren't even getting credited. Apparently there is no contract for them so they are not obligated to stay or keep quiet from what I have seen from ex anilivers. They are cheap and horrible. They would have been better off giving them a character design and made it on a premade customizable model. An ex liver that left before debut showed their art before and after it was put in the ai program and it looks real and normal before and fucked up and ai after. It's really sad I already didn't trust this app when it came out.
It's not even rigging. This is like flash animating, if you never took into account the lessons most flash aimators back in the day had to learn.
@@RobertStoll Yeah it seems something like that then but done by an ai program definitely looks bad because it's not human made. I think it would be more acceptable if they were more transparent because they are just scamming a bunch of people who signed up at this point.
❤❤
Heyaa!! Ex-anilive jumpstart here! I have used AniLive since Jumpstart 1. I had quit being on of their jumpstarts and deleted the app. I had been going live for months without knowing about this. When my sister also applied for AniLive, she was given a VTuber model but guess what? It looked exactly like mine, I mean not COMPLETELY but REALLY similar, this made me so upset and frustrated because I had work for a WHOLE year on my vtuber model so no way i'd let them do that! I went live and talked about it to a few people(a few ppl idk since I barely got any attention or viewers) And at this point of the time, Im siding with the new jumpstarts who got AI-looking generated model. This is literally unfair and I feel bad for them, I had created a video about this but im currently re-doing it because my voice was cracky and hoarse. I feel bad for the VTuber artist who worked/who was hired for/by Anilive, they really should compensate them because why would they take advantage(I don't know how else to word this so..) of them? That's literally just cruel, I can relate to the artist that she was naive because I was too.
Reminds me of the website/free Steam game that let you "make" 3d vtubers and they were all stolen paid models from Booth and Gumroad. On their official website they were even reselling the models and acting like they owned them
istg there's so much drama abt AI in 2024
As someone really interested in illustration, the moddel at 18:30 would look really nice to actually draw
10:32 so theyre basically just copying iriam. In every way 😰
5:50 - i would be surprised, if the company would even win a single % of a lawsuit at this point. At least in the EU contracts can be voided in specific cases and the company intending to ruin the VTubers reputation is quite a strong argument for that.
If anything, they could have commissioned artists, but could be using AI for editing/transitions. But the fact that they won't name the artists tells volumes.
i already felt weirded out when i got a DM from them a month ago trying to get me to apply to their program, and now this..
15:24 albedo from overlord
16:56 navia
youtube algorithm be crazy, i keep getting your video recommended to me.... keep on going kat~!
I used to use anilive when I was sick and throwing up I had a comfort streamer too everyone there was so nice :(( this is so sad
recently i was looking threw casting call club and i was looking and they where making people audition for this compony
You know what they look like? Those tiktok ai filters, like the ones that will make you 'look like a disney princess' and it just slaps it over your face while you talk and move
Honestly weather its ai or not its still really crappy as someone who is an artist but doesn't make models or riggs its still pretty bad ngl. again even if its not ai its still scummy to give people who are most likely newer to vtubing a model that just has that obvious of issues i get that its free but still.
edit: finished the whole vid and point still stands however i also wanna add its just decitful to give someone a quickly made model that doesnt even have the same functions of a normal pngtuber
This absolutely could have been a case of AI animation on real illustrations. AI animation is really janky still and absolutely would add those AI traits to the art because it makes use of generative AI. I don't think that's any better than using AI art honestly, though. And we haven't seen any still illustrations without the artifacts yet...and they were hiring gen AI engineers
They invited me to the jumpstart program and I thought it was weird they required your model's files if you already had one, now that I know they use AI I can tell why. They were trying to train their shitty AI with it LMAO. Big red flags so thankfully I didn't join/give them my files
AI is making me crash out... 👊😡
FR
You can’t really punch AI since it’s well digital like the closest you will get is punching a robot and even then you would get pain by the metal
@@ma1-t1f lol
The Live2D rigger in me is screaming watching this. A human would NEVER look at this and say "eh good enough". There's a reason Live2D models are expensive and this is a great example of what it looks like when you try to cut costs.
Also I remember drama with this app in the Live2D community some time ago. They asked for the mco3 files (editable project files in Live2D) from every VTuber who signed up which is a massive concern because it means they can temper with the rig and use it as a template. Most riggers don't give out project files for that reason, we don't want our hard work getting messed with or copied.
oh, how funny, this was the same company that had a controversy about how their app operated
idk if it still works that way now, but basically, you had to UPLOAD your model files to their website, and the way they do the tracking of your movements is by sending the tracking data from YOUR phone to THEIR servers
this differs highly from how vtubers usually do it because these two things are done all on your computer, so this raised concerns about them maybe using the model files for AI training, and also privacy concerns from people thinking they uploaded your camera image to their servers
which also meant that they don't livestream VIDEO output to viewers devices... they apparently stream the model and tracking data directly! which means that when someone opens a vtuber stream, their model gets DOWNLOADED to their device, and someone could theoretically steal it if they know how!! isn't that fun?? /sarc
edit: fixed typo
Me when I don’t do research
Here's the thing: They're advertising themselves as a VTuber app, not a PNGTuber one, or some general streaming service. Sure, they mentioned getting a free ANIME CHARACTER, not a VTuber model, but it doesn't make it any better, because they specifically emphasize on VTubing. And while we're at it, the animations and rigging are HORRIBLE. Not only are the models probably a mix of AI and very poor adjustments, but the movements themselves don't make any sense whatsoever. It looks so bad and I'm so sorry for the people that received THIS as their models to debut with. I'm so sure it's not what they were promised, or what they believed they would receive, and it can very well be considered false advertisement or deception
Now, how can they be so sure that the "illustrations" aren't AI? They're either lying, or they were scammed into believing that the images are legit. They don't have any means of knowing if the people they supposedly commissioned didn't or did use AI. All of this is so sketchy
I mean, that's what happens when a company claims to give you something for free. If they're giving away "free models", then the quality will be representative of that, and as we can see, it is very much bad
this video was just made to churn out content, wasn't it? zero research, the same three false claims over and over again, and no regard for what the actual streamers were saying.
don't leave this up. it's blatant misinformation, really mean towards the actual artists that drew those avatars, and every single vtuber affected by the app is in these comments telling you that this isn't the main issue about this app.
also please just stop "analysing" ai stuff. you keep fumbling more and more with each video, where you call out stuff that isn't even an issue with ai anymore and confuse it with human mistakes or, in this case, with a different way to animate that doesn't use rigs. you also don't see to have an issue taking displate money, when they are so riddled with ai images - so ig you only care if you can make money off of it.
Let's leave out the "rigging" part of that feature. The art is still clearly AI. Not only that but there are amazing png model that look incredible and way better than that slop. Like they could've commissioned an actual PNG artist for a good price.
its like in between png tubing nd vtubing fully wit the rigging, like someone jus day 1 started rigging their models wit improper seperations for the art 💀 its choppy as u can pause every single frame and its so stiff when it sways.
these vtubers simply wanted an opportunity but was given this nonsense and not what they potentially deserve.
i cant even say they should keep their design of their models that was given, because it was most likely stolen through the ai too.
This is an example of a company coming in and literally sucking the fun out of a culture
Most of model is cheap, but its not AI (yet), it has the same beginner mistakes I've made while drawing.
The closest one to AI is the purple dress girl with the dress being held like a royalty, but its not yet conclusive. You can call it cheap or rushed or lacked quality control, but not yet AI.
You should have put the Generative AI Engineer hiring part from the LinkedIn on the front because that's the meat of the issue.
18:15 is this a fully rigged model? What is it the half vtuber Half pngtuber model if it’s fully I don’t believe it’s made by AI because AI is too stupid to make a fully riggable model but if it’s a rigged png model I guess that would be possible
It doesn't look like genAI art, more like that kind of dancing filters where you can stick your picture and they make your photo dance. I think they're just scrimping on money tbh. No one is just gonna give out free live2d models. Also if you're a company/org the monthly fee you need to pay to Live2d/Cubism is really high loll. Using this kind of cheap filter is how they're giving 'models' out for free.
As for the artist/crediting, nijisanji allows the artists to claim 'mama'-ship but A LOT of their old stock (they have a huge stash of models that they just pick out to use and distribute) do not have mamas and are under nijisanji's ownership. They just pay artists to draw designs and rig it in-house back then. In Japan they rarely credit the artists, esp in those 2d card gacha games, unless you're a really famous/popular artist, then they'll plaster your name everywhere and call it a collab lol. A lot of your fav JP artists are doing contract work outside (ever notice they disappear for months then come back and start drawing your couple again?) but they are never credited or have NDAs. I am in no way defending Anilive, just discussing how the real art industry works outside of twitter vtubers and artists lol
After this video I got an ad for adobe's ai stuff 💀
They tried to be like IRIAM, who did p well in jp and are current expanding to the West, but anilive's tech is too far behind
These models are definitely quantity over quality
Ohhhhhhhh, not Albedo and Zero Two… please tell me it’s a joke.
man, anilive sounds great but its such a downer if they give models like this, like if you dont wanna go through the process of making a legit model due to lack of money and time or just being lazy, offer a PROPER non-AI PNG-tuber model to start them with, that's still technically a Vtuber model AND its much easier and quicker to produce
I feel so bad for the person who got this model and rig as their prize, at least it was free i guess...but this is just a scam
WHAT HOW DOES THAT WORK
Say you want to animate an eye blinking.
For the best effect, you could draw an open eye and a closed eyelid. Then layer the closed eyelid on top of the open eye whenever the eye was "closed." This is how most vtubers nowadays are rigged; it's like an intricate digital puppet with layered parts that can be effected by the face tracking.
If you wanted to cut a corner, however, you could just take that open eye and draw and note the area around the eye, along with the color of the eyelid. When the eye closes, color in that area with the color of the eyelid, and when it is open don't color it in. That's what these guys are basically doing. Flash animators used to do the same, only it looked really stilted unless you put effort in and, well, cut your model up into multiple pieces.
If it was real art, not ai. Why wouldn't they say names? Understand artist want to maintain anonymous but kf people are saying your art is ai i imagine they would come forward
I mean if my art was being hated on and called ai I doubt I'd take credit for it. What an insult! Why give people the ability to hate it to my face?
@WordyDird_art I understand, personally as a artist I would hate go have people calling my art ai
"We do not use AI" Source: Trust me, bro
Not even Vtuber with this “rigging” lol just basically a PNGtuber
I WAS ANTICIPATING YHIS BIDEO!! even though I'm just a vtuber clipper I was thinking of making it myself , lmao
I put this in another comment's reply but this app (not their Jumpstart program) is a nice alternative for people who already have an OC and have made art for it but don't know how to rig or have not found an artist and/or rigger within their budget to commision an actual Live2D model. The obvious quality issue with an actual just normal PNGtuber model(? is that the right word? im not sure) with proper blink art and/or expressions is obviously going to be noticed. But maybe the artist is charging an ungodly amount just to make another art piece that's basically the same just with closed eyes and/or different mouth shape.
We've already been using AI face recognition in our daily lives a lot anyways: from Face ID to beauty filters and other fun filters used in social media apps. Not everything is bad with AI. AI was originally meant as a tool to help humans anyways, not take over our jobs. Like this app will not be taking over any rigger's job because of the obvious quality difference. Except for actual AI ART that's been generated from an AI Art Program since that's actually stealing from artists as their training data. It's unfortunate art theft has been a problem for millennia.
Stealing people's faces is still stealing. If you want to use someone's face for anything, you pay them.
@RobertStoll People's faces? Wdym?
I kmow Anilive! They asked me to join them, i said nah
Reality is far better 😭 (they use 3D models)
ive made my ownnnnn screw thatt
I was scared it was Hyper Online lol. I’m happy it wasn’t them.
Ok
Hey kat
YALL ARE FAST WHAT
Wow
I really hate how ai bros are trying to convince artists they need to intergrate ai into their art or the “will be left behind” I remember a tracer whos art i was investigating because they traced before you could look at it and see ai traits that they didn’t correct. (Unfortunately they’re a massive comment deleter and delete comments that don’t agree)
Naw
crazy
YAYAYYAYA
Vtuber model ≠ Pngtuber model. Looks like that App is for Pngtuber models with limited rig on eyes and mouth (Made with neural networks maybe?) Also first models looks like hand made not AI. But you need to take in consideration that maybe are just concept art for the model that was just rigged for pngtubing.
Not everybody have 700USD for a full rigged Vtuber model and some just could use this for curiosity and not a "necessity".
Isn't ai wayyyy better than these models wth lol
😂
Are there companies like this out there that aren't so.. controversial? I only hear about these company ideas that I love and would want to support, on drama channels.. And they only get on drama channels once they've done something wrong🥲