I know you think this is taking forever, but I’m honestly blown away by the amount of progress you’ve made in such a short period of time and with minimal assistance from anyone else. The level of quality that you’ve achieved so far is professional grade. I’d happily wait several years to see an undergrads movie of this quality, though hopefully it doesn’t take that long lol. All this to say, keep up the great work 👍
I noticed you had a green background in the 3D software. If you're using this to do a "green screen" colour key, don't do it. Just render with the alpha channel and use a pure black background to avoid any fringing when dealing with multiplied vs pre-multiplied alpha.
For all the people dunking on you for using AI, I have to say that you should push ahead. This is how you make AI worth while. Making it serve the artist. It looks better to me then any toon shader I've seen so far, and I've seen a lot of impressive stuff in Blender. I also think that the high frame rate it a bit weird looking, but it's also just a floating head, so it's hard to tell at this point. Making a full demo with a full character will give a better impression of how it looks in motion. Maybe going to a lower framerate will help, but perhaps embracing the oddities of the medium is the key. Who knows. I'm glad someone is looking at this seriously, and finding out how it works, and how it doesn't.
To be fair he's using his own art to train it and he's not requesting the AI to create art from scratch. So he owns the entirety of it. It's good to see that AI is being used ethically in here. Probably people are tired of hearing the word and forgetting why it was a problem for them.
@tom901ful if he is using AI generation software, and he trained on TOP of it, yeah, thats making results that look like the training data, but theres still MILLIONS of stolen art and images included in the base training of the AI, and that cant be removed, because AI image generation must have up to millions of images to work, and while he did put his own art into the later training, if he had trained on a completely empty ai image model, it wouldn't look as good, because i doubt he drew millions of his own drawings to train the AI on. This is literally just using the exact same ai which has stolen training data in it, and just pushing it in a direction to make art that looks like the images he gave it, it is stills built upon millions of images artists didnt consent to be used to train AI
@@VJArt_No true do a bit more research. If using ethcially designed control net and loras there is no stolen data. An no stolen images used to generate images. Issue lies with the LAION-5B data used to improve the current version. It's like saying my driving instructor learns how to drive by stealing someone's car. So, since he thought me how to drive, there are ethical implications on me driving. Instead, of the driving instructor stealing vehicles.
@@chriscandide643 Loras are seperate training done on top of pre existing image generators. It is just training the RESULTS to look like the traning data in the lora, but all the stolen data is still in the original ai generator, and the original ai generator couldnt be this good without so much theft. Loras only make the results look more like the new training data, but the process is still based on theft in the original AI generator. If you can run the ai generator without the lora, than its built upon stolen data. its not like the lora overwrites all the original traning
@@VJArt_ incorrect and please read my original response correctly. "Control net" should have read " Checkpoints " You don't have a full understanding of how Stability AI's stable diffusion works. I mentioned Checkpoints and Loras, if both are user generated or ethically trained there is no issue or using stolen data. The primary concern with stable diffusion and were people split hairs is the initial training using LAION-5B dataset. Analogy, you learn information from an individual that's gains their knowledge through ellict means. You the individual full trained not commiting any crimes but your mentor has. Doesn't make you the individual unethical but the person who trained us wrong. ( From and end user stand point, not speaking legally ). When using the software in its current state, ethicality isn't black and white. Stable diffusion is modular concerning both checkpoint and Lora model sets.
OMFG GENUINLY THIS IS SO FUCKING COOL LIKE I GENUINELY LOVE THE WAY YOU'RE CORRECTLY USING AI Like the love and work into this is just a chefs kiss, I'm so excited to see this movie that's in the works
Yeah, but some frame reduction should definitely take care of that. The motions feel too smooth because there are too many frames. Action is correct, movement is correct, there just needs to be less inbetween frames I think.
This is one of the more ethical uses of AI in artwork that I've seen. There are severe budget and workload restraints, and it still has critical human involvement in polishing the finished product. So far I think it's pretty impressive! You may need adjustments in the source 3D animation to make sure it's adequately snappy, I find that often this 3D-to-2D stuff winds up looking way too floaty, losing some of the expressive snap/jerk in a lot of cartoon movement. The rigging you showed for the exaggerated facial expressions is promising, showing you've still got some good instincts for the way the finished product should feel.
Animator here! I'd love it if some squash and stretch was added. It's something HYPER utilized in 2D, and would definitely help sell the 3D/2D look. Not sure how well the shader would take it on, but I reckon it's worth a shot.
Even a short and quickly thrown together video detailing that there was some kind of setback and there hasn't been substantial progress because xyz would at least be something. These things do happen, and while some will probably be unhappy with that, I think most of us would at least appreciate the honesty. Still want to see this movie get done someday. By all means, take the time needed to get it done, but when you say you want to keep your fanbase updated monthly, it's best to at least mostly follow that, even if the news isn't always good. I don't claim to necessarily know the difficulties entailed in getting a work like this completed, and maybe I'm baselessly optimistic, but even if there are updates that aren't quite as full of new things as you are we might like them to be, as long as there is a clear forward momentum most of the time, I don't think too many will have issue. There may even be good entertainment to be had in tales of things going wrong, as long as they don't dominate your content. I did find substance to your tales of having issues with pupils and mouths and such being made to work with our animation method, for instance. Anyway, all the best to you, Pete, and team. Undergrad was a favorite of mine in my last couple years of high school, so whatever happens, I genuinely hope you're doing ok, movie or no movie in the end. ... the Duggler?
I have no doubts anymore than using 3d models was the right choice. It looks really good. The only hint left that this is not made frame by frame is the smoothness of the animation. If it's possible to lower the framerate to something comparable to the show then it would look spot on.
Looking good. I have a few thoughts however. - I'm sure lighting will be a separate pass. - For characters with unique headshapes like Gimpy, You may need different models depending on where they're looking, IF you're wanting to be faithful to the original style, which it seems so. I just mean that breaking the models will probably be a must. - I know these are tests and not at all representative of final animations but I think for 2D styled characters like this, you don't want them to look "tweened" like in 2D animations, so more solid poses and lesser framerate the better I feel, but we'll see when you get there. I'm not actually familiar with the show, or really the style at all, but sometimes smoother looks cheaper. Very interested in this project and hope to see more videos like this :)
This looks so good. Though I feel like using AI in this case might be a bit overkill. 3D software these days can produce very good cel shading right in the software, which would require a lot less clean up. Blender's "Grease Pencil" tool is a great example, because it literally draws 2D lines right on top of the model render. So it kind of actually is 2D. Using those techniques would save a ton of render time, and reduce all the trial and error. But hey, this pipeline looks pretty damn good, so I can't complain.
I've never seen Undergrads, this behind the scenes series just popped up in my recommendations. I think these one of the most interesting animation process videos I've seen. It's very cool to see you problem solve with a limited budget and resources. Good luck and i'm very excited for the movie, even if I havent seen the show yet haha
Can I please get the 3d models! 🙏🏽 I make VR content and would love to port these into VRchat so we can not only show everyone who loved undergrads its revival but also so that we can show more people and have more people accept this type of animation. PLUS, I know so many people who would love to be gimpy or cal in VRchat. Why deprive us?? 🤷♂️
Blender would be easier: grease pencil line art for outline, defuse node to rgb into a colour ramp node for colour+shadow colour material and then a camera and light
The tufts of hair under his hat seem to grow and shrink almost at random, with one visibly shrinking into nothingness near the end of the clip. The brim of the hat also "flows", or has some other weird property I don't have a name for. The rest looks pretty good, though! Love the creative compositing solutions.
You missed the point so very badly on this, its about companies and studios replacing actual artists with cheap AI, Williams is using ai for what its actually meant to do, serve the artist and aid, not replace.
5:24 Dude what are you talking about! You passed in A MONTH from having to draw all pupils to this amazing Nitz's face, that by the way looks better than I tought it would ever be without drawing it yourself To be honest, a month ago, I didn't think it was gonna look that good, I tought "I could watch an Undergrads movie that looks okey I don't need it to look perfect". But dude, this really looks cool, I'm very impressed
It's so unfortunate I just recently found out about and missed the kickstarter, I totally would've bought the $300 collectors bundle.. even if I'd end up broke. But keep up the great work on the movie, it's great to see a beginning to this project. It feels almost surreal seeing the characters being reanimated.
I'm glad you're putting out these videos because with the animation industry in a slump, this shows a workflow that indie animators can use to make high quality content on their own. It also shows studios and broadcasters the very real limitations of AI workflows; they're not magic and still need talented artists to make anything good with them. Hopefully we see a few indie stars come out of this like in the games industry.
i do want to mention those hair strands on Nitz were kind of looking wonky but besides that the animation looks great, I'm not the type of guy to trust ai in the animation process but i think your the first person to utilize ai in the animation process correctly
pete williams please 😭 i know it takes forever to do stuff but we would all appreciate even a quick message to just let us know you're still working on it
Yeah usually the whole point of using motion capture is that it captures lots of micro-movements that make things look realistic, but realism isn't the goal here. Rendering out the animation at a lower framerate would probably help, but the actual animation needs some tweaking too.
As an animator who does only 2D for all my animations frame by frame old school style this is kinda why lol 😂 all that would drive me insane. It's nice to have full control of a character all by hand and being able to draw their mouth in anyway I want without worry of how it'll look for the most part Even though i know most animation anything has gone strictly math and technical more than just hands on like it used to be. No hate of course it's always impressive but I'll always be more stunned by hand drawn anything 😅
Interesting, but why did you use AI here? This is not a problem AI is needed to solve (and as you saw, got in the way), you can calculate the outlines yourself, and manually set edges to be outline, as well as doing it per object in a model, giving you more artistic control for *less* effort (and no, I"m not talking about using the builtin-software you discussed earlier). No need to specialize eyes and mouth, no weird junk going on. Heck if you wanted to, you could synthezize curves directly from the outline, and apply warble, line thickening, or what ever 2d stylistic effect you're trying to go for. I guess it might be hard if you don't know programming or math... but a team kind of needs to know how to write shaders and understand normals etc to be doing any kind of professional 3D art at all. If you wanted to use AI, you should have started with 2D and used it for tweening to drastically cut down on line art and inpainting (and drastically simplifiying the process for the AI to learn). But if you're hell bent on using 3D rigging and trying to translate that to 2D, you're just going to bleed resources and waste time compared to just creating shaders yourself to produce the effect you want.
I think they mention in the first part that the goal is to mimic the actual style of the original show and having the main cast be able to believably stand with supporting cast in the same scenes, while cutting down on cost. Doing 2d puppet animation would change the vibe drastically, needing multiple rigs or simply defaulting back to traditional animation (which would be extremely expensive and time consuming), and using a solely 3D solution would end up looking unmistakably 3D in the way toon shader stuff looks like if they do not invest in super high quality rigs and, again, expensive animation. They are probably working with a tiny budget and this approach is more time consuming at the beginning, but if it pans out it can increase their possible output substantially as well as the perceived 2D quality of the end result. The 3d models they use here can be more basic than a full fledged cinematic production would require and the AI shader would ideally fix and manage the visual complexities of how the 'physics' of the characters would work and render it closer to the 2D style, allowing for animators to have the benefits of 3D animation (ease of tweaks, previzualization, yadda yadda). There's a lot of good possibilities in this method, honestly looking forward to more about it.
@@erostashnot to mention it’s not like they’re not just going to some rando “AI Animation” site and typing “Undergrads, nitz, movie, face animation.” It’s their own model, their own art, their own animation.
Two things I noticed just watching this video: 1. Parts of the animation after conversion seem to warp or disappear. For example, at 5:07 you can see on the left, the hair tuft under the brim of the hat sort of looks like its moving in and out like some sort of hair tentacle, and then at 5:12 on the right side, part of the tuft seems to just slide out of place. This is almost certainly an issue with whatever AI process is involved, and forgoing AI entirely to just do it all with shaders would eliminate that issue. (And that's on top of the obvious AI wobbling) 2. As far as making things look more 2D, it's gotta be framerate. 3D animations always seem to forget about this when trying to emulate 2D animation, and the end result is just too smooth, almost a bit uncanny. 2D animation is done frame by frame with timings all laid out, so cutting out some frames and animating on something like 2s would definetly help out with the choppier 2D look, though keeping things on 1s can of course be done for certain actions.
One thing I really suggest is editing/deleting frames and such to make it look like it’s animated on ones and twos. Flat animation looks odd when it is always consistently on ones
For context, animations are usually 24 FPS Animating on 1s: each frame is different, usually used with action scenes, emotional scenes, or complex scenes. Visually looks like 24fps Animating on 2s: only changes visually every other frame, used during dialogue, basic movement, background movement. This is used the most in most animations. Visually looks like 12fps
It would need to be extremely though out,I think this won't actually work because this is all animated with mocap,there isn't thought out extremes,secundarys and in betweens,without those a proper process can't be done,since really your idea would mean taking the extremes and secundarys and duplicating their length and eliminating some of the in betweens,the issue being that we don't know what the invetweens are,and taking time to see every animation frame by frame to pass it through a process of eliminating frames isn't acceptable considering that a lot of time is already being spend on the toon shading,also it's just isn't necessary,lost of animated movies in 2d are animated in 1's examples being all the animations of James Baxter,actually speaking most 2d movies are animated in 1's
@@JoseArmandoSaint-HilaireAcosta For some programs like blender, there are settings that allow it to animate in 2s automatically with it's tweening, which would apply to mocap as mocap is just inputting the same data someone would be inputting by hand, but through tracking.
@@myheromedia-verse You could probably still add some 2d effects on top after the initial 3d to make it feel more 2d. Combining minor 2d assets with 2d-styled 3d looks amazing. Hi-Fi Rush does that a lot and it looks spectacular.
...so how about those monthly updates. It doesn't need to be something grand and super edited every time, though the last few were pretty cool. Just talking at the camera for a few minutes about how things have been going.
It… it is what it is, and is also impressive what you managed to do, but when you start designating roles to hire you might as well hire 2D animators because it just doesn’t get there
@@ELTABULLO art is art. If the creator of the show wants to use 3d and use 2d to add flair, that’s his choice, and there’s nothing lost in doing that. It’s absurd to gatekeep this stuff when the question is really: is it good or is it bad? Not “hurr durr, should’ve been 2D, I can’t believe the studio would make a creative choice over the one I think is best”.
@@ELTABULLO “impressive what you managed to do, but when you start designating roles to hire you might as well hire 2D animators.” To give you the benefit of the doubt, I could assume you that you mean his style of animating the 3D models isn’t there, and that if he had some team more equipped for the 2D look he’s going for, he could make it better. This may be the case, but the gatekeeping, if not, lies in the rejection of the medium itself. To deny any one medium credibility is gatekeeping what art is and should be.
This was a show that meant so much to me as a kid for it being that perfect blend of realism and cartoon level humor, I’m so excited to see the final results when it’s finish it’s looking amazing:)
Why not let us fund you to make Undergrads by selling an in-depth tutorial on this pipeline? I'd be willing to drop cash to help fund you if I could buy step-by-step on this process.
This is so absolutely fucking wonderful to see advance! Take all the time in the world man! Im shocked nobody picked the movie up but at the least once this baby launches and you get the distribution youll keep a way bigger piece of the pie :)
This was such a fun cartoon for me back when I was younger. Looking forward to seeing the movie. But, I gotta say, I hope it ends with some kind of glimpse into their futures so I know how they all turn out...
its pretty cool! but uh i feel like there is a ton of issues with this mainly with..well all of the issues you came across, while it looks good i dont think the ai is being very...well useful here
I commented on the previous video about how I thought that the 3D to 2D stuff wouldn't work since the 3D models would make everything too perfect compared to the process of 2D hand animation. After this video I still think this sadly. Its very impressive, the whole process for making the 2D style animation with the 3D models, but it still looks like a 3D model to me. It looks just like you're tweening a 3D model but overlaying a filter to make it look 2D. Now, I'm no 3D animator so I wouldn't know the best advice to give to make it look more hand drawn however I do want to just throw out an idea that maybe would help. I think if you were to do multiple animations of the 3D model but each animation had slight variations to the tweening, then put that through the process of making it 2D, and then spliced together frames from each separate animation it might make it look more 2D. I base the idea on how stop motion is shot because I feel stop motion and 2D hand drawn animation share a similar feel in that both have forms of movement that are imperfect. Depending on the amount of frames a stop motion movie is shot in, the stop motion animator will have to judge where a limb or body will be based on feel. This leads to imperfect placements of said limbs but perfect enough for our brains to fill in the in between frames making it look seamless. Maybe if you applied a similar line of thinking to your process, intentionally creating imperfections for the viewers brain to fill in the gaps, it would look less toon filter and more hand drawn. I'm no genius though, I just didn't wanna give criticism without providing a possible solution, hope your movie turns out as good as you'd like it though!
Funnily enough though I might add, a 2D to 3D process of animation actually does work to keep both of the best qualities of each form of animation. If anyone is curious go check out the movie Klaus to see what I mean, really pretty and phenomenal movie. It still looks like a 2D hand drawn animation but with an insane depth and cool lighting effects of 3D. It was made with 2D hand drawn animation and put through a process to apply lighting and texture that makes it look 3D.
This looks so, so much better; really think you should've pushed that this was where you were heading in the last video. Worthikids on YT has shown me how much can be done with this 3D 2D art style, I still can't believe their whole channel was done in Blender.
Exciting! . I'm very impressed with all of your effort to replicated all of the details to get it as close to the original hand drawn animation. I can see this will definitely be worth the wait. Thanks again for posting all of the updates keep up the great work team.
Accidentally stumbled upon this video. First I thought it was some test character, then saw the card "undergrads the movie" and then it clicked for me. I only saw when it back when I was a kid, it was on 2x2, a russian tv channel, barely saw two episodes and then it disappeared from my life. Now it went to a strange weirdness, like, fully rediscovering something you barely saw but can remember the art style with small tinge of reminder. Now I'll check the original out. thank you
A lot of people would be suggesting Blender due to its 2D animation tool suite. Its very strong to combine 2D and 3D workflows in one software. Blender was used in the pipeline for Spiderverse, particularly for the sketchy effects on The Spot. Blender has a whole 2D animation engine built into it as well. A lot of people would be suggesting Blender because its the 3D software they'd know. Its the right price (free) for the average person, so they know it better. Its got a number oddities that dont suit a production environment. Even used in Spiderverse, Sony programmers had to write proprietary extentions to make it behave the way they wanted. Spiderverse was mostly animated in Maya, because of course it was. If you've got a pipeline that works, use it.
it sounds like you just made way more work for yourself by opting to use AI for this. the AI lines are all jiggly and weird. why not just use the 3d software toon shader…
Oh man, the progress is great! The animation is looking fantastic, it really looks like it's coming together. The waits are definitely not a problem, seeing monthly updates on new Undergrads stuff is wild and awesome
I strongly believe you could have used a outline for the whole model and in some areas use an emission shader, Blender has showed us the power to create 2D looks in 3D, for example across the spider-verse were made with Blender
i love that i found this youtube channel. it's really interesting seeing how esp the creator of a tv show approaches the creation behind their own project. it's very personal to also get to see and get shared these tips for other creatives. cant wait to see what more cool stuff there is in the future!
Love hearing all the "boring technical stuff"! Take your time, superb work from everyone involved! Just some framerate tweaking and de-smoothing needed to get that better 2D feel. While animated movies tend to get bigger budgets, which means smoother animation, it still doesnt give that butter-smooth feel like 3D. Everything looks amazing, and I'm super excited to get a tutorial for creating 2.5D characters/shaders! BUMP for the algorithm! (Though please, please focus on good, solid writing as much as possible. If there's anything Bluth movies taught me, it's that good animation can only take you so far. The true heart and soul is in writing and characters. Shite writing - shite movie, no matter how long anyone slaves over visuals. Just look at Good Dinosaur.)
![[Live] ศึกมหึมันส์มวยไทยสังเวียนเดือด เวทีมวยชั่วคราวสวนสยาม | จันทร์ 30 กันยายน 2567](http://i.ytimg.com/vi/pM1k6f7JXUc/mqdefault.jpg)
I know you think this is taking forever, but I’m honestly blown away by the amount of progress you’ve made in such a short period of time and with minimal assistance from anyone else. The level of quality that you’ve achieved so far is professional grade. I’d happily wait several years to see an undergrads movie of this quality, though hopefully it doesn’t take that long lol. All this to say, keep up the great work 👍
I think once he has all the tools in place it should move faster
People don't really know how long animation really takes
I noticed you had a green background in the 3D software. If you're using this to do a "green screen" colour key, don't do it. Just render with the alpha channel and use a pure black background to avoid any fringing when dealing with multiplied vs pre-multiplied alpha.
Please consider releasing the soulless cut when the movie is done.
Eyes and mouths are overrated
That would be hilarious
For all the people dunking on you for using AI, I have to say that you should push ahead. This is how you make AI worth while. Making it serve the artist. It looks better to me then any toon shader I've seen so far, and I've seen a lot of impressive stuff in Blender.
I also think that the high frame rate it a bit weird looking, but it's also just a floating head, so it's hard to tell at this point. Making a full demo with a full character will give a better impression of how it looks in motion. Maybe going to a lower framerate will help, but perhaps embracing the oddities of the medium is the key. Who knows. I'm glad someone is looking at this seriously, and finding out how it works, and how it doesn't.
To be fair he's using his own art to train it and he's not requesting the AI to create art from scratch. So he owns the entirety of it. It's good to see that AI is being used ethically in here.
Probably people are tired of hearing the word and forgetting why it was a problem for them.
@tom901ful if he is using AI generation software, and he trained on TOP of it, yeah, thats making results that look like the training data, but theres still MILLIONS of stolen art and images included in the base training of the AI, and that cant be removed, because AI image generation must have up to millions of images to work, and while he did put his own art into the later training, if he had trained on a completely empty ai image model, it wouldn't look as good, because i doubt he drew millions of his own drawings to train the AI on. This is literally just using the exact same ai which has stolen training data in it, and just pushing it in a direction to make art that looks like the images he gave it, it is stills built upon millions of images artists didnt consent to be used to train AI
@@VJArt_No true do a bit more research. If using ethcially designed control net and loras there is no stolen data. An no stolen images used to generate images. Issue lies with the LAION-5B data used to improve the current version. It's like saying my driving instructor learns how to drive by stealing someone's car. So, since he thought me how to drive, there are ethical implications on me driving. Instead, of the driving instructor stealing vehicles.
@@chriscandide643 Loras are seperate training done on top of pre existing image generators. It is just training the RESULTS to look like the traning data in the lora, but all the stolen data is still in the original ai generator, and the original ai generator couldnt be this good without so much theft. Loras only make the results look more like the new training data, but the process is still based on theft in the original AI generator. If you can run the ai generator without the lora, than its built upon stolen data. its not like the lora overwrites all the original traning
@@VJArt_ incorrect and please read my original response correctly. "Control net" should have read " Checkpoints " You don't have a full understanding of how Stability AI's stable diffusion works. I mentioned Checkpoints and Loras, if both are user generated or ethically trained there is no issue or using stolen data.
The primary concern with stable diffusion and were people split hairs is the initial training using LAION-5B dataset. Analogy, you learn information from an individual that's gains their knowledge through ellict
means. You the individual full trained not commiting any crimes but your mentor has. Doesn't make you the individual unethical but the person who trained us wrong. ( From and end user stand point, not speaking legally ). When using the software in its current state, ethicality isn't black and white. Stable diffusion is modular concerning both checkpoint and Lora model sets.
OMFG GENUINLY THIS IS SO FUCKING COOL LIKE I GENUINELY LOVE THE WAY YOU'RE CORRECTLY USING AI
Like the love and work into this is just a chefs kiss, I'm so excited to see this movie that's in the works
It's like a reverse Klaus and I love it! Instead of being a 2D movie that looks 3D, it's a 3D movie that looks 2D.
Honestly it looks too smooth, just barely noticeable as 3D under a 2D "filter"
Yeah, but some frame reduction should definitely take care of that. The motions feel too smooth because there are too many frames. Action is correct, movement is correct, there just needs to be less inbetween frames I think.
This is one of the more ethical uses of AI in artwork that I've seen. There are severe budget and workload restraints, and it still has critical human involvement in polishing the finished product. So far I think it's pretty impressive! You may need adjustments in the source 3D animation to make sure it's adequately snappy, I find that often this 3D-to-2D stuff winds up looking way too floaty, losing some of the expressive snap/jerk in a lot of cartoon movement. The rigging you showed for the exaggerated facial expressions is promising, showing you've still got some good instincts for the way the finished product should feel.
now THIS is a good way to use AI
Definitely lower the framerate or remove every other frame and smooth it out. It's too real.
This is the first thing ever that vaguely reminds me of Klaus. That movie was great.
Animator here! I'd love it if some squash and stretch was added. It's something HYPER utilized in 2D, and would definitely help sell the 3D/2D look. Not sure how well the shader would take it on, but I reckon it's worth a shot.
Even a short and quickly thrown together video detailing that there was some kind of setback and there hasn't been substantial progress because xyz would at least be something. These things do happen, and while some will probably be unhappy with that, I think most of us would at least appreciate the honesty. Still want to see this movie get done someday. By all means, take the time needed to get it done, but when you say you want to keep your fanbase updated monthly, it's best to at least mostly follow that, even if the news isn't always good. I don't claim to necessarily know the difficulties entailed in getting a work like this completed, and maybe I'm baselessly optimistic, but even if there are updates that aren't quite as full of new things as you are we might like them to be, as long as there is a clear forward momentum most of the time, I don't think too many will have issue. There may even be good entertainment to be had in tales of things going wrong, as long as they don't dominate your content. I did find substance to your tales of having issues with pupils and mouths and such being made to work with our animation method, for instance.
Anyway, all the best to you, Pete, and team. Undergrad was a favorite of mine in my last couple years of high school, so whatever happens, I genuinely hope you're doing ok, movie or no movie in the end.
... the Duggler?
I have no doubts anymore than using 3d models was the right choice. It looks really good. The only hint left that this is not made frame by frame is the smoothness of the animation. If it's possible to lower the framerate to something comparable to the show then it would look spot on.
Keep the low frame rate honestly, it looks fantastic. Reminds me of the new animated Spiderman movies and older Saturday morning animations
Looking good. I have a few thoughts however.
- I'm sure lighting will be a separate pass.
- For characters with unique headshapes like Gimpy, You may need different models depending on where they're looking, IF you're wanting to be faithful to the original style, which it seems so. I just mean that breaking the models will probably be a must.
- I know these are tests and not at all representative of final animations but I think for 2D styled characters like this, you don't want them to look "tweened" like in 2D animations, so more solid poses and lesser framerate the better I feel, but we'll see when you get there. I'm not actually familiar with the show, or really the style at all, but sometimes smoother looks cheaper.
Very interested in this project and hope to see more videos like this :)
In video games we typically use blend targets or morph targets for facial expressions. Facial rigs don't normally provide enough detail.
This looks so good. Though I feel like using AI in this case might be a bit overkill. 3D software these days can produce very good cel shading right in the software, which would require a lot less clean up. Blender's "Grease Pencil" tool is a great example, because it literally draws 2D lines right on top of the model render. So it kind of actually is 2D.
Using those techniques would save a ton of render time, and reduce all the trial and error.
But hey, this pipeline looks pretty damn good, so I can't complain.
The final composite of Nitz is Awesome!!!
This is how I’m finding out there’s an Undergrads movie coming out. I’m stoked haha
I've never seen Undergrads, this behind the scenes series just popped up in my recommendations. I think these one of the most interesting animation process videos I've seen. It's very cool to see you problem solve with a limited budget and resources. Good luck and i'm very excited for the movie, even if I havent seen the show yet haha
All episodes of the show are on youtube.
take your time. when the movie is done itll be your proof of work and may revolutionize alot of the animation industry.
Gimpy would approve of this tech
Exactly I could see Gimpy making his own Star Wars movie this way.
this like that 1 adam sandler christmas movie
😂im not alone it looks like 8 Crazy Nights
Can I please get the 3d models! 🙏🏽 I make VR content and would love to port these into VRchat so we can not only show everyone who loved undergrads its revival but also so that we can show more people and have more people accept this type of animation. PLUS, I know so many people who would love to be gimpy or cal in VRchat. Why deprive us?? 🤷♂️
lol my 1st thoughts too.
Translation: “I’ll PaY YOu iN ExPOSurE!”
This is definitely a better use for AI
Waiting less time between updates, even if there's no progress, would be nice
This reminds me of the curious George movies
Try lowering the framerate to make it look like it was aninated on 2s
Looks like the Curious George movie's style
Blender would be easier: grease pencil line art for outline, defuse node to rgb into a colour ramp node for colour+shadow colour material and then a camera and light
The tufts of hair under his hat seem to grow and shrink almost at random, with one visibly shrinking into nothingness near the end of the clip. The brim of the hat also "flows", or has some other weird property I don't have a name for. The rest looks pretty good, though! Love the creative compositing solutions.
Yeah no this looks really quite good
It looks clean
You're doing an incredible job, guy. Too many ungrateful people complain about how long it takes, I'm super appreciative of these updates.
I genuinely am liking the style! It's really starting to look like the old show in a new way!
I don't get the people bitching about AI when he's using his own art for it
Some people just want to play intellectual nothing to see here
Literally like I heard he’s making a tutorial at the end of the video and I’m so damn excited.
You missed the point so very badly on this, its about companies and studios replacing actual artists with cheap AI, Williams is using ai for what its actually meant to do, serve the artist and aid, not replace.
@@wikwayerAnd you're not one of those.
Yeah this is the best way to use it too. Home grown data made with purpose.
You should upload these models to the Garry's mod workshop and for Left 4 Dead 2 as a mod
Is that Adam Sandler twelve crazy nights
Literally what i was thinking 😭
It’s been over 20 years I think we can wait a little longer. With that being said… Can’t wait to see it!!!
5:24 Dude what are you talking about! You passed in A MONTH from having to draw all pupils to this amazing Nitz's face, that by the way looks better than I tought it would ever be without drawing it yourself
To be honest, a month ago, I didn't think it was gonna look that good, I tought "I could watch an Undergrads movie that looks okey I don't need it to look perfect". But dude, this really looks cool, I'm very impressed
You know if this guy was smart, he would sell this technique to other people in studios. Did she produce animation more cheaply and quickly.
Pete next update dude cmon
It's so unfortunate I just recently found out about and missed the kickstarter, I totally would've bought the $300 collectors bundle.. even if I'd end up broke. But keep up the great work on the movie, it's great to see a beginning to this project. It feels almost surreal seeing the characters being reanimated.
do you know why they aren’t accepting anymore money?
I'm glad you're putting out these videos because with the animation industry in a slump, this shows a workflow that indie animators can use to make high quality content on their own. It also shows studios and broadcasters the very real limitations of AI workflows; they're not magic and still need talented artists to make anything good with them. Hopefully we see a few indie stars come out of this like in the games industry.
undergrads MOVIE 💜 please UPDATE 💜
i do want to mention those hair strands on Nitz were kind of looking wonky but besides that the animation looks great, I'm not the type of guy to trust ai in the animation process but i think your the first person to utilize ai in the animation process correctly
pete williams please 😭 i know it takes forever to do stuff but we would all appreciate even a quick message to just let us know you're still working on it
I’m thinking that tutorial should take a backseat to the film, let the film be the priority then do the tutorial as like a DVD special feature
Man, I have infinite respect for the ceaseless work you put into that project. Thanks for the update!
Having be so smooth looks odd, but promising. I feel like it's a bit "over animated" and doesn't have the original charm. Too many movements
Yeah usually the whole point of using motion capture is that it captures lots of micro-movements that make things look realistic, but realism isn't the goal here. Rendering out the animation at a lower framerate would probably help, but the actual animation needs some tweaking too.
As an animator who does only 2D for all my animations frame by frame old school style this is kinda why lol 😂 all that would drive me insane. It's nice to have full control of a character all by hand and being able to draw their mouth in anyway I want without worry of how it'll look for the most part
Even though i know most animation anything has gone strictly math and technical more than just hands on like it used to be.
No hate of course it's always impressive but I'll always be more stunned by hand drawn anything 😅
both ways are really hard in my opinion, requires a lot of knowledge and patience
Interesting, but why did you use AI here? This is not a problem AI is needed to solve (and as you saw, got in the way), you can calculate the outlines yourself, and manually set edges to be outline, as well as doing it per object in a model, giving you more artistic control for *less* effort (and no, I"m not talking about using the builtin-software you discussed earlier). No need to specialize eyes and mouth, no weird junk going on. Heck if you wanted to, you could synthezize curves directly from the outline, and apply warble, line thickening, or what ever 2d stylistic effect you're trying to go for. I guess it might be hard if you don't know programming or math... but a team kind of needs to know how to write shaders and understand normals etc to be doing any kind of professional 3D art at all. If you wanted to use AI, you should have started with 2D and used it for tweening to drastically cut down on line art and inpainting (and drastically simplifiying the process for the AI to learn). But if you're hell bent on using 3D rigging and trying to translate that to 2D, you're just going to bleed resources and waste time compared to just creating shaders yourself to produce the effect you want.
I think they mention in the first part that the goal is to mimic the actual style of the original show and having the main cast be able to believably stand with supporting cast in the same scenes, while cutting down on cost. Doing 2d puppet animation would change the vibe drastically, needing multiple rigs or simply defaulting back to traditional animation (which would be extremely expensive and time consuming), and using a solely 3D solution would end up looking unmistakably 3D in the way toon shader stuff looks like if they do not invest in super high quality rigs and, again, expensive animation. They are probably working with a tiny budget and this approach is more time consuming at the beginning, but if it pans out it can increase their possible output substantially as well as the perceived 2D quality of the end result. The 3d models they use here can be more basic than a full fledged cinematic production would require and the AI shader would ideally fix and manage the visual complexities of how the 'physics' of the characters would work and render it closer to the 2D style, allowing for animators to have the benefits of 3D animation (ease of tweaks, previzualization, yadda yadda). There's a lot of good possibilities in this method, honestly looking forward to more about it.
@@erostashnot to mention it’s not like they’re not just going to some rando “AI Animation” site and typing “Undergrads, nitz, movie, face animation.”
It’s their own model, their own art, their own animation.
Question, does anyone else see David Duchovny in Nitz? He always looked like he was the cartoon version of Fox Mulder.
Two things I noticed just watching this video:
1. Parts of the animation after conversion seem to warp or disappear. For example, at 5:07 you can see on the left, the hair tuft under the brim of the hat sort of looks like its moving in and out like some sort of hair tentacle, and then at 5:12 on the right side, part of the tuft seems to just slide out of place. This is almost certainly an issue with whatever AI process is involved, and forgoing AI entirely to just do it all with shaders would eliminate that issue. (And that's on top of the obvious AI wobbling)
2. As far as making things look more 2D, it's gotta be framerate. 3D animations always seem to forget about this when trying to emulate 2D animation, and the end result is just too smooth, almost a bit uncanny. 2D animation is done frame by frame with timings all laid out, so cutting out some frames and animating on something like 2s would definetly help out with the choppier 2D look, though keeping things on 1s can of course be done for certain actions.
One thing I really suggest is editing/deleting frames and such to make it look like it’s animated on ones and twos. Flat animation looks odd when it is always consistently on ones
For context, animations are usually 24 FPS
Animating on 1s: each frame is different, usually used with action scenes, emotional scenes, or complex scenes. Visually looks like 24fps
Animating on 2s: only changes visually every other frame, used during dialogue, basic movement, background movement. This is used the most in most animations. Visually looks like 12fps
Animating on 3s also exists, but it’s usually limited to lower budget shows; or very very tiny background characters and scenes that you won’t notice.
It would need to be extremely though out,I think this won't actually work because this is all animated with mocap,there isn't thought out extremes,secundarys and in betweens,without those a proper process can't be done,since really your idea would mean taking the extremes and secundarys and duplicating their length and eliminating some of the in betweens,the issue being that we don't know what the invetweens are,and taking time to see every animation frame by frame to pass it through a process of eliminating frames isn't acceptable considering that a lot of time is already being spend on the toon shading,also it's just isn't necessary,lost of animated movies in 2d are animated in 1's examples being all the animations of James Baxter,actually speaking most 2d movies are animated in 1's
@@JoseArmandoSaint-HilaireAcosta For some programs like blender, there are settings that allow it to animate in 2s automatically with it's tweening, which would apply to mocap as mocap is just inputting the same data someone would be inputting by hand, but through tracking.
@@themanysirs1814 but it isnt comparable to proper anymation in 2's,it just cuts the frames and looks like a game when its laggy
this is so neat brah
Was enjoying these videos quite a bit, and you got a ton of views on this one. Shame to see it end.
Ok you promised to update every month and now it's been two months and more
I think at some point, it might be easier just making a really good toon-style 3d model.
honestly yeah, like drawing frame by frame just takes way more time and honestly hurts the wrist more, I would know cuz I'm an animation student
@@myheromedia-verse
You could probably still add some 2d effects on top after the initial 3d to make it feel more 2d.
Combining minor 2d assets with 2d-styled 3d looks amazing.
Hi-Fi Rush does that a lot and it looks spectacular.
@@Ze_N00B Adding this to my notes, thanks dude 😄
The ratio for your personal dedication to making it streamlined and cost efficient for your budget is astounding. Shout out to you nitz-guy
If Clone High can be brought back, so can Undergrads
...so how about those monthly updates. It doesn't need to be something grand and super edited every time, though the last few were pretty cool. Just talking at the camera for a few minutes about how things have been going.
You just gotta hope the ai stays at least 95% consistant
Looks like that one Adam Sandler animated chriatmas movie
That movie was terrible, but tbh that animation was pretty good
@@thewiseoldfoxyeah
please update 😞
Looks awesome! Telltale Games did a great job also of making 3d into a 2d inked/painterly style
It… it is what it is, and is also impressive what you managed to do, but when you start designating roles to hire you might as well hire 2D animators because it just doesn’t get there
No.
Gameplays 😭😭 ok PewDiePie, calm down
@@ELTABULLO art is art. If the creator of the show wants to use 3d and use 2d to add flair, that’s his choice, and there’s nothing lost in doing that. It’s absurd to gatekeep this stuff when the question is really: is it good or is it bad? Not “hurr durr, should’ve been 2D, I can’t believe the studio would make a creative choice over the one I think is best”.
@@TheOnlyPedroGameplays Gatekeep how?
@@ELTABULLO “impressive what you managed to do, but when you start designating roles to hire you might as well hire 2D animators.”
To give you the benefit of the doubt, I could assume you that you mean his style of animating the 3D models isn’t there, and that if he had some team more equipped for the 2D look he’s going for, he could make it better.
This may be the case, but the gatekeeping, if not, lies in the rejection of the medium itself. To deny any one medium credibility is gatekeeping what art is and should be.
Thanks again for the update. Im really enjoying watching this process.
This was a show that meant so much to me as a kid for it being that perfect blend of realism and cartoon level humor, I’m so excited to see the final results when it’s finish it’s looking amazing:)
Why not let us fund you to make Undergrads by selling an in-depth tutorial on this pipeline? I'd be willing to drop cash to help fund you if I could buy step-by-step on this process.
I wanna see a cartoon with this animation!!!
(Is this the same technique used for X-Men 97?)
I wouldn't mind seeing the movie as an animatic.
This is how I find out about there being a movie? This is exciting
It's been years and I'm still excited for this project.
This is so absolutely fucking wonderful to see advance!
Take all the time in the world man! Im shocked nobody picked the movie up but at the least once this baby launches and you get the distribution youll keep a way bigger piece of the pie :)
This was such a fun cartoon for me back when I was younger.
Looking forward to seeing the movie.
But, I gotta say, I hope it ends with some kind of glimpse into their futures so I know how they all turn out...
University Admin: Mr. Walsh, your financial aid has been suspended.
Blind, mouthless Nitz: I have no mouth and I must scream!
This looks SO SICK!!! Awesome work man, thank you for these updates!
I really miss old MTV
The Head. The Maxx. Aeon flux.
YES! we're that much closer!
its pretty cool! but uh i feel like there is a ton of issues with this mainly with..well all of the issues you came across, while it looks good i dont think the ai is being very...well useful here
like with some of this issues it feels like you could already do this FASTER if you just do it manually instead of doing this whole debacle
With more progress update videos, i noticed youre getting more and more traction!
Very happy to see this upload and thanks for sharing!
I commented on the previous video about how I thought that the 3D to 2D stuff wouldn't work since the 3D models would make everything too perfect compared to the process of 2D hand animation. After this video I still think this sadly. Its very impressive, the whole process for making the 2D style animation with the 3D models, but it still looks like a 3D model to me. It looks just like you're tweening a 3D model but overlaying a filter to make it look 2D.
Now, I'm no 3D animator so I wouldn't know the best advice to give to make it look more hand drawn however I do want to just throw out an idea that maybe would help. I think if you were to do multiple animations of the 3D model but each animation had slight variations to the tweening, then put that through the process of making it 2D, and then spliced together frames from each separate animation it might make it look more 2D. I base the idea on how stop motion is shot because I feel stop motion and 2D hand drawn animation share a similar feel in that both have forms of movement that are imperfect. Depending on the amount of frames a stop motion movie is shot in, the stop motion animator will have to judge where a limb or body will be based on feel. This leads to imperfect placements of said limbs but perfect enough for our brains to fill in the in between frames making it look seamless. Maybe if you applied a similar line of thinking to your process, intentionally creating imperfections for the viewers brain to fill in the gaps, it would look less toon filter and more hand drawn.
I'm no genius though, I just didn't wanna give criticism without providing a possible solution, hope your movie turns out as good as you'd like it though!
Funnily enough though I might add, a 2D to 3D process of animation actually does work to keep both of the best qualities of each form of animation. If anyone is curious go check out the movie Klaus to see what I mean, really pretty and phenomenal movie. It still looks like a 2D hand drawn animation but with an insane depth and cool lighting effects of 3D. It was made with 2D hand drawn animation and put through a process to apply lighting and texture that makes it look 3D.
This looks so, so much better; really think you should've pushed that this was where you were heading in the last video.
Worthikids on YT has shown me how much can be done with this 3D 2D art style, I still can't believe their whole channel was done in Blender.
Yeah Ian is the gold standard for quality 2D looking 3D animation as far as I'm concerned
Reminds me a lot of the way that Tycho and Brock are modeled in the Poker Night at the Inventory games
the guy who voiced max in the cartoon voiced the little fat kid too
Oh now I get it from other updated we saw. The other one from few years ago. I hope they put that in theaters.
6:35 i still think you should view diff animations pple have done with blender that look 2d
Maybe a Patreon so we can fund your noodles and bills while you make our dream come true?
Exciting! . I'm very impressed with all of your effort to replicated all of the details to get it as close to the original hand drawn animation. I can see this will definitely be worth the wait. Thanks again for posting all of the updates keep up the great work team.
2:50 Nitz has entered the Avatar State.
I am super stoked to watch this when it comes out in 2053😂
Accidentally stumbled upon this video. First I thought it was some test character, then saw the card "undergrads the movie" and then it clicked for me. I only saw when it back when I was a kid, it was on 2x2, a russian tv channel, barely saw two episodes and then it disappeared from my life. Now it went to a strange weirdness, like, fully rediscovering something you barely saw but can remember the art style with small tinge of reminder. Now I'll check the original out. thank you
Holy shot I’m just baffled someone besides me is thinking about Undergrads
That’s not just any someone, that’s the original creator lol
A lot of people would be suggesting Blender due to its 2D animation tool suite. Its very strong to combine 2D and 3D workflows in one software. Blender was used in the pipeline for Spiderverse, particularly for the sketchy effects on The Spot. Blender has a whole 2D animation engine built into it as well.
A lot of people would be suggesting Blender because its the 3D software they'd know. Its the right price (free) for the average person, so they know it better. Its got a number oddities that dont suit a production environment. Even used in Spiderverse, Sony programmers had to write proprietary extentions to make it behave the way they wanted. Spiderverse was mostly animated in Maya, because of course it was. If you've got a pipeline that works, use it.
Awesome stuff, can't wait to see more!
it sounds like you just made way more work for yourself by opting to use AI for this. the AI lines are all jiggly and weird. why not just use the 3d software toon shader…
Most toon shaders on the market are eh… they are okay but not great, which is why I’m so hyped with this new cel shading pipeline.
Oh man, the progress is great! The animation is looking fantastic, it really looks like it's coming together. The waits are definitely not a problem, seeing monthly updates on new Undergrads stuff is wild and awesome
I strongly believe you could have used a outline for the whole model and in some areas use an emission shader, Blender has showed us the power to create 2D looks in 3D, for example across the spider-verse were made with Blender
Klaus was also done in Blender but that movie is 2D with some 3D elements
but AI is future. It can do anything that can already be done but wors with more steps
@@candlewick8363 ai is for lazy people
That test animation looks amazing. So convincing too.
i love that i found this youtube channel. it's really interesting seeing how esp the creator of a tv show approaches the creation behind their own project. it's very personal to also get to see and get shared these tips for other creatives. cant wait to see what more cool stuff there is in the future!
Love hearing all the "boring technical stuff"! Take your time, superb work from everyone involved! Just some framerate tweaking and de-smoothing needed to get that better 2D feel. While animated movies tend to get bigger budgets, which means smoother animation, it still doesnt give that butter-smooth feel like 3D. Everything looks amazing, and I'm super excited to get a tutorial for creating 2.5D characters/shaders! BUMP for the algorithm!
(Though please, please focus on good, solid writing as much as possible. If there's anything Bluth movies taught me, it's that good animation can only take you so far. The true heart and soul is in writing and characters. Shite writing - shite movie, no matter how long anyone slaves over visuals. Just look at Good Dinosaur.)