Why my 3D animated NPR project doesn’t look 2d enough?
Вставка
- Опубліковано 27 жов 2023
- When you are working on an NPR still-frame, everything is fine, but then, when the camera or your character moves, a lot of 3D alarms shout out and the trick is revealed.
Shading, camera animation, parallax, shadows, frame-rate are few of several aspects that will be mentioned on this presentation, going through every pipeline department and explaining how to avoid 3D feeling on your stylized 3D animated project.
Take risks, explore new workflows and do something that makes a difference, instead of wondering (despite having bought the ultimateNPRrebornuniversalshader) why my 3D animated NPR project isn't 2D enough?
"Why my 3D animated NPR project doesn’t look 2d enough?" by Toni Mortero --
Learn more about Blender Conference 2023 at conference.blender.org/2023/
#BCON23 #b3d - Фільми й анімація
That skull animation blew my mind. It has the energy and fluidity of a 2D animation with the level of detail only realistic for a 3D model. Absolutely stunning.
He used Keymesh addon for the animation
@@Skedooosh i need to get that. that is gonna be awesome
Yeah, adding noise to all the little things did wonders.
Tho my personal tastes say I would have preferred the end result to be on 1's still. Smooth 2d looking animations are rare, due to budget reasons.
And later on I get why they'd be rare on NPR too for budget reasons. Cause a lot of it involves going over frames, swapping models mid action, and other tricks.
Would like to add my two cents, with alot of 3D animations that look like 2D, one thing i see all the time is with the perspective. Its always perfect, between two characters or even different scenes with the same character its extremely difficult to get them to match. So one thing that might be a fun thing to do is to maybe have as multipass render where characters are rendered either with different cameras set to different focal lengths, or to somehow procedural distort them in the z axis relative to the camera. Anyways great talk alot of helpful insights in here.
Agreed! That's what I've noticed as well. Perhaps Geometry nodes could handle this? Procedurally distorting based on the distance from the camera or how far away the normal is pointing from it (maybe).
I agree that it would help simulate 2D animation better. However I think that it is better to retain some of the 3D's accuracy. Since the entire point of emulating 2D art in 3D is to make use of the perfect perspective and cameras in order to produce scenes with better cinematography easily, or to achieve accurate and automatic foreshortening etc...
Thats so cool
@@jutraim2422just make sure it's not distracting
I think like rendering characters with a transparent background then animating the scene where that 2d render is on a plane might achieve that. Allow you to do more parallax effects and make it less perfect.
Thank you everyone for showing some love to my presentation!
I just want to add some credits that I completely missed during the talk (too nervous!)
- This personal project that I'm currently working on is a fan art based on this comic book called "Solo" (by Oscar Martin) @_oscar_martin_
- The fight shot is animated by the amazing Alan Carabantes (@alancarabantes)
- 2D blood animations are made by the great Emanuel Espinoza (@luna.mmee)
Also, thank you to the Blender Conference for giving me the chance to share all of these things with you.
that was amazing! I was always wondering what is happening with the anime scene, like, the Berzerk remake... It was so bad. Now I see clearly what was the problem! They were not thinking outside the box like a true gamer! Great work!
@skilifavas4016 hahaha thank you!
I think it's kind of a natural thing/issue in 3D workflows, an expected behavior. You start letting your computer make bad decisions for you and to accept what your viewport gives you in return.
May be 3D tools can improve your production times, but as a 3D artist you have to change your process and kind of lowgrade your "quality expectations" in order to get rid of 3D slavery hehe break the game!
🤩
Thank so much for this presentation Toni! You inspired me to really learn Blender 😉
i think part of the issue is the density of design. 2d animations are often made relatively simplistic in design to minimize how much work is needed per frame. some high density npr designs look 3d because of the detail even when they are still. of course designs are subjective and this isnt always the case
really good talk, i will say another thing is a selective use of frames is another thing that makes it look 2d. Animators usually switch from animating in 2s to animating in 4s to depict changing speeds and to make animation easier.
Yeah, it is tempting for new animators to blanket animation with 2s or 3s or 4s. But when I do study frame by frame on Sakugabooru and the likes, the lesson comes very simply. You simply draw the frames that your animation needs and you have time for. So with the masters the frames per seconds vary wildly, often inside the same shots they will go from stills of seconds per frame right down to animating on ones when fluidity is needed and budgets allow. And you can cheat framerates by selectively cascading the cadence to make animation on twos and threes feel like ones because there is new movement in each frame.
Also. Detail levels on masters shots adapt wildly as well, both to make drawings faster and to keep the animation readable at playback.
Sakugabooru is amazing study material. Also, I see a lot of people animate the background and characters all on twos instead of separating them.@@jmalmsten
how he even made the horizon floor move according to camera. It is clear that the workkflow includes drawing on top of the 3d camera, correcting the 3d model, render, and correcting it again in 2D. Yep, indeed there's a lot of 2D work on top of that cool 3D outline and fills. Great expo!
Because of things that go deeper than the model’s surface, like muscle tension and relaxation. In CG, it takes a lot more fine tuning and knowledge of the 3D program to make the character look and move in a detailed and believable way. With 2D animation, you individually illustrate everything and have way more control over each frame. Not just elasticity in terms of cartoony exaggeration, but in terms of realistic and detailed muscle movements that you would typically only notice up close. Amazing stuff can’t wait to see where 3D goes from here
In my opinion, asking "Why try to make 3D look 2D when you can just draw?" is like asking "Why buy a car when you can just walk?".
Am I mistaken in thinking that NPR 3D will be, if not easier, than at least less time consuming than traditional animation?
Speaking as a hobbyist on and off: it depends. In traditional animation, what you see is what you get. You draw what you want, and if something needs to be done that isn't set up yet, draw that, too. So depending on the project, more time on the actual making of the thing. Something like Loud House with the floating mouths would take less effort, something like Wakfu with the epic battles would take more effort. That's comparing 2D to other 2D, though. It's still object-based if you're using software, but it has a different set of limitations and a different distribution of timesinks.
In 3D, there's a bigger timesink towards setting up. Making your 3D models, checking the shaders, how to do the linework if any... from what I've seen, it still involves drawing, primarily to plan ahead. If there's a need for an expression and your characters can't do it, it's way harder to add in 3D than it is in 2D. But on the upside, it can be easier to propagate that change to other characters. Something like Paw Patrol, where a lot of the pups look like they have the same topology and probably the same rig, can take advantage of that. Something like Digital Circus, where all the characters are vastly different, cannot.
Also, one answer to that question a lot of people might forget: not everyone can draw, and learning 3D can help you get maybe not the same skills in drawing, but at least help you get to definite results much faster. It is possible to do a 3D render of something you cannot draw, and then rotoscope it or ask animators to draw following that render as a reference. That's how the Big Ben scene in Great Mouse Detective was done, and at least one combat scene in Ranking of Kings. There was too much going on on-screen in both cases for traditional animators to worry about (turning cogs and falling debris along with camera movement, respectively), so 3D rendering gave them a reference to work towards.
Again, not an expert, so take it with a grain of salt, but that's what I've gathered so far.
I desperately want to know how he did that geometry nodes head/face replacement. That was insanely cool.
Spectacular talk. You are an amazing artist!
This is really incredible and sparks so many ideas
GREAT TONI!!!! ❤🔥🔥🔥🔥🔥🔥🔥
This is so interesting! Thank you for the presentation.
AMAZING Talk!!! Thanks a lot for your knowledge!!
@tonimortero Thank you so much for sharing your work and ideas! Fantastic presentation. I would be first in line to purchase any courses you create. Would love to learn more!
really cool talk! thanks for this! :)
Amazing artist and best person!
Looks amazing!!!
¡Muy útil...!!!! Buena presentación.
Really impressive, congratulations Toni
this is ridiculously great!
I think to add even more imperfections, adding extra bones to the rig in order to move elements of the face around, changing the scale of certain bones for certain frames and maybe even drawing the shading by hand on flat characters would sell the 2D effect a lot more.
He showed off doing that at 14:27. And the rigging stuff he shows off at 12:20.
Amazing adaptation of the Oscar Martin comic, thank you for addressing this topic and including in the presentation that Tatsuiki Tanaka Akira scene he animated, it is my favourite and i'm also reading a book writen by him at the moment.
I'd like to see that full rat man animation
We are gonna have some great iconic games in the future with these techniques, I'm going to study UE5 and Blender just for this even though i'm in mid 30's. I used to make Flash animations back in 2006 but quit as it was too much work.
Nice!
wow!
Woah what was that Bakugo/Inosuke piece??
Wow
The title hits right where it hurts.
🤩
interesting
Ranxerox 👍
NPR with 2D inspiration as well as mixing 2D and 3D can be a way of doing stuff that is possible in 2D but that would require too much work. In fact, that's what 2D does by using a lot of 3D backgrounds. Mixing 2D and 3D is beautiful
The king of NPR :)
Something I've noticed as well when studying this: our brains seem to be able to pickup on the polygons in the geometry. Primarily on the boarders of curves like elbows and shoulders. I'm guessing this is a similar observation to seeing "the pores stretch" in CG faces.
well, irl pores stretch, we don't get new pores to fill the wider gaps...
So I dont know much about creation of styalized 3D, but whenever I see stuff like that I just think that they always overdo it with the outline making it very thick and not scaling it down as characters get closer to the camera. Because of that the outline looks static and... well 3D, like a black chunk sticking out of the character. Same with outlining hard shapes and too much details like teeth and such, maybe using second simplified wobbly teeth model (like just two white blocks instead of individual teeth) may help
I don't think fps limit or camera restrictions are a good thing. They generally make the visual experience worse, and can only be taken in specific circumstances (like spiderman being awesome in every pose mid-seing, and that lets us cherish each longer).
What really breaks the 3rd wall is how shading moves - if it follows the shape of the object so well that you see the 3d shape, then you can't unsee it. Outline can be any, but shading is what sells and kills it. Decrease the angle range acceptable for lower shade color and you get the 2D look. Balance the colors. Remove standard smooth gradient shading. That's the greatest impact.
My man just hits on the right spot💀
God he is making me so anxious
Can we puhleaaase have an even lower volume? It's not like we're here to listen to anything important on topic that's totally common. Right?
I think that this 2d styled 3d is probably the best and most ethical use case for AI. We know 2d animation studios are brutal work environments so I feek like if you could animate 3d scenes, pull out key frames every 30+ or so frames and then hand draw them then use style transfer to match the rest it could really bridge the gap. One of the biggest issues with 3d is the uncanny perfection so in a way the slight inconsistencies of style transfer would introduce more "human"-esque variation. If you're using hand drawn frames by your studio as the style reference youre also avoiding the big AI art issue of stealing, you'd only train it on in house data sets. Even with 2D animation you could use it to give foreground elements the extra painted detail of traditional static backgrounds. And then it would drastically reduce animator workload (I know 3d animation isn't easy but CGI animation is popular because it can be much less work intensive than 2d), saving animators from brutal conditions while allowing gorgeous consistent 2d style animation to be made.
It seems like he was forced to talk on the stage
*Why Doesn't My 3-D Animated NPR Project Look 2-D Enough?
*Non-Photorealistic Rendering
Great talk, but one thing he fails to notice, 2d artists simplify shapes and details as much as they can, a lot of 3d 2d style stuff doesn't do that. Even the Suzanne sketch is too noisy for any animator to have drawn it. U get that detail in manga and comics, look at cartoons and anime, that's a rare level of detail.
because its not 2D?
every fucking year there has to be some production issues.
now i have to blast the damn volume way up because nobody could be bothered to check the audio output. so now heaven forbid i get a notification or play something else and blow my ears out.
😂😂😂
tf2 pyro