Notice: Purchasing, Trials and Accounts Have Changed. What You Need To Know: Part II >>

Topic - NPR using new cel edges and motion blur

Page 1 of 2
   1  2  Next  Last 
Hello!

A while ago I came up with ( http://forums.newtek.com/showthread.php?t=33453&highlight=Sketch ) a way in LW to create some NPR effects like these early experiments:











The basics of it are outlined in this old tutorial I wrote:

http://www.oocities.org/gbreazeal/LW_Sketch_Tutorial_1.pdf

In a nut shell, along with animated procedural textures, it uses a noise procedural to deform a mesh very quickly using fractional frames in combination with motion blur and an ink line shader (though it can use any surface).

It seems modo does not evaluate displacement at render time for motion blur (at least in my tests so far). That said, I'm thinking that using the new deformers would probably work as they are (hopefully) calculated at rendertime with motion blur.

OK off to test (when I get time). Maybe someone else already knows the answer to this displacement/motion blur riddle?
I kind of have it working with a transform deformer that has an animated scale, with a radial falloff that is using a noise procedural (that is animated).

Not the best but getting there...I'm too new to the modo render engine and how it handles motion blur to know where to tweak first. There is no step option...only distance etc. for blur. At least the main part is "working". Probably should have used a more interesting model!

First modo sketch line style attempt

If I could find a way to deform out along the normals instead of just what looks to be a global push out (tried local cord on the transform deformer...doesn't seem to change much) that might help.

The camera blur is tricky...66% gets a deformation blur...75% little to none etc....it's a little random.

Any suggestions would be welcome!

Message edited by Gideon Klindt on 3/2/2012 - 6:27 PM

very interesting look ..
as a matter of fact .. those must be the best example of npr that i have ever seen .

subscribed :-)
Wow, please keep trying. It looks very interesting.
Have you tried the displacement trick using multiple frame passes?

B
That was actually one of my next NPR thoughts to try Brad! I have seen it done that way in different packages- it is a reliable NPR technique.

Adding some variance to the line width directly would help...

Plenty to play with to get some NPR- the main thing here is getting shorter lines with some fade and blur to create a gestural look vs ink. Does the motion vector pass detect displacement changes? That could be a post solution.

Thanks again for the great release Brad (and the lux crew and beta testers)!

Plenty of room to play and get some NPR looks going with the new shaders and pass system, even if I can not replicate my LW tech verbatim.

Message edited by Gideon Klindt on 3/3/2012 - 5:59 AM

Hmmm. I don't mean render passes I mean the per frame passes. It is a channel on the render item. If you set it above 2 modo will render the frame multiple times with subframe offsets and blend them together. This way you can get the texture to motion blur I think. It is more similar to the stepped LW blur.

B
Quote from BradPeebler :
Hmmm. I don't mean render passes I mean the per frame passes. It is a channel on the render item. If you set it above 2 modo will render the frame multiple times with subframe offsets and blend them together. This way you can get the texture to motion blur I think. It is more similar to the stepped LW blur.

B


Sweet baby cheesus!* No wonder I couldn't get reliable results- this is enirely dependent on sub frames for non jittered line animation. Ok, need to keep my noob thinking hat on (and RTFM more) and keep plugging away- I had almost given up on this method in modo...will investigate- thank you Brad!

PS- what the heck are you doing on the message boards after the launch on a Sat night? You a) have a very understanding family b) a great work ethic c) must be more into 3D apps than most developers d) be obsessed with your own creation. I can only say thanks to all of those- and how amazing is it that the prez of lux is watching over his flock of users- just awesome- thanks!



*sweet baby cheesus is a registered trade mark of JavisJones Inc.
Not meaning to burst someone's bubble, BUT I don't think modo's frame passes supports animated texture parameters just like motion blur doesn't support animated texture locators and I think this applies to displacement. You'd have to do it via deformers by using the displacement as Falloff Value. Here is a sample Scene.

Yazan
Quote from Yazan Malkosh :
Not meaning to burst someone's bubble, BUT I don't think modo's frame passes supports animated texture parameters just like motion blur doesn't support animated texture locators and I think this applies to displacement. You'd have to do it via deformers by using the displacement as Falloff Value. Here is a sample Scene.

Yazan


No bubbles bursted yet! That is the method I have been testing- using deformers and falloffs with a procedural. I look forward to looking at your scene and seeing what you might be doing differently.

It is understandable that modo does not blur textures and displacements as they would just add a lot of unneeded overhead for most rendering situations. Maybe it will be a non supported channel some day?!

Still- I find it odd that frame blending would not blend textures to some degree- though I now doubt it will do it like it is handled in LW.
Thanks Yazan for the file. I was over complicating things a bit in that I didn't know you could use the deformer falloffs directly with the FX part of the rendertree. I thought you had to add a transform deformer first etc.

Thank you Brad as well for pointing out the render passes channel- that is getting me much closer.





Like I said, these are closer, but I'm finding controlling the noise is kind of hard compared to if it's displacement. Mainly...it seems biased towards one axis- anyone know why this is? I also just can't seem to get it to effect the lower part of this mesh much. I've tried moving around the texture locator, and different projection settings besides solid (spherical etc.) and while it changes things...it seems something is up with this area around the neck- any suggestions on how to get a more even deformation around the mesh? This problem becomes more pronounced the smaller the noise is...maybe I should try another noise?

The second image is with the mesh not sub divided. It creates disjointed polygon artifacts, but if you lower the line quality you can get a nice, almost random stipple effect with the frame blending blur going on.

I think when I get more control over the noise I might then try and get more over all control by using weight maps on areas. Places like the eyes could be aided with the use of a weight map to either kick up the effect or lower it. Using and incidence angle gradient on the noise might help too...if we can do that with FX tree shading...?

I'll keep at this when I have time (uh..."real work" for clients calls...even though this is more fun). Thank you again Brad and Yazan for the help!

Message edited by Gideon Klindt on 3/4/2012 - 10:40 PM

OK...gotta quit and do work oh and eat!

One more...kind of paint/clay like. Would have to add several features to get it "right". Looks like surface shading can't be animated on sub frames (you were totally right Yazan) so...gotta find a work around to some of my old tricks...

Just another little doodle...



- T.
Hi guys... cool to see others trying this.
I'll try to share my attempts from the beta sessions...

If I recall correctly, the main thing (even with the frame passes) is that the blur is not incremental, but smooth.
Result looks like a smear. Also - dither worked really well, and is unavailable. It gave a feeling of substrate texture.

That was on my list before I got swamped here... testing a camera projected texture. But... a true npr system got me sidetracked ;-)
TomW- yeah it's quite funny actually! I never use this for anything really, but I do use other NPR techniques for clients. Those are a little less...trying to emulate "traditional" media if you will though!

David: How is that project coming along? You're working closely with a developer on an stand alone NPR solution correct? If it's the one I'm thinking of it looks awesome! Any plans for when that will be out?

@Tom: that's so not funny.

@wacom: things are so incredibly thick. Have so many interested projects that are gung-ho up to the point where they should be signing the dotted line. So many that I think we HAVE to be on the breaking point. But also so many that we've begun working on ways to take things into our own hands.
I'll try to post some new things soon. Lots of framework dev, so the images themselves don't look up to snuff.
Thanks for this thread Wacom!

I too tried this in modo and haven't found a way yet to get close to the results in Lightwave.
The main thing missing seems to be a way to not have the renderer interpolate/dither the motionblur.

The rendertimes for the sampled cell edges are also a bit forbidding - that's the one cool thing in Lightwave, the superfast edges... ;-)
For my testing, I used good old incidence instead ;-)

I also tried to build a stepped sub-frame curve with as many steps as I set the Frame Passes to, but still the renderer does interpolate and smear the result.

Well, for the time being it seems I'm back in Lightwave for things like that... :-)

Hey David, wow, that NPR project of yours seems in the making for ages. I hope I'll live to see it! :-)

Cheers,

Tom
Yeah, of all the things I missed about LW was it's "dumb" edge and line mode- that really wasn't so good for making NPR on it's own, but great for fiddling with. They were blazing fast- that helps a lot. That and the way the motion blur worked with sub frame animation. Again, not the kind of blur you want to use for most animation- but for this purpose it was perfect.

I've been too busy on a project to fiddle around with this, and do a few mini-tutorials (like just sampling the BG for "BG GI" etc.)- lots of cool new things to try and jack in modo though so I look forward to when I get some down time.

I figure between you, david maas, and I we should be able to see what can be done in Modo in terms of NPR hack'n though! Looks like a lot of people are getting good results just using the tools straight up- but nothing really sketchy or painterly yet- getting closer though eh?

Looking forward to your shaders BTW Thomas. In your free time, if you come up with a node system that fits in the layer system of Modo I'll be first in line to buy! I can dream right?
Hahaha - yeah, that would be cool ;-)

I can't see me do that in the foreseeable time though and I guess it would be a lot of work with an uncertain future - in the end, we want nodes in modo itself, implemented natively, don't we ;-)

What could be done right now AFAIK would be a meta-shader that opens an external GUI with nodes - something like the first node plugins for Lightwave before it went native.
Since all available nodes in this shader would have to be coded for it specifically (you couldn't use the modo shaders and textures AFAIK), it would be a lot of footwork for a limited result.
And if Luxology should come up with some kind of node system for shaders, it would be garbage.
This would only make a certain amount of sense if Lux would definitely say that they don't plan shadernodes for the next decade or so ;-)

But yeah, I think I could like this too!

As far as the NPR goes, I so far haven't found any hack that circumvents the interpolation, and as you say, it's flaws make the Lightwave Renderer perfect for this.

Three examples from Lightwave from me - maybe we can inspire somebody at Luxology to implement a no-interpolation mode and an option for sub-frame-displacement evaluation or even "dumb" fast edges...? :

First a crazy-text experiment for a film title - the whole thing evolves slowly in the animation - subframe displaced 2point-polys:


And two other tests I did - sorry, for the stupid test object - to see how this this kind of rendering works for stereoscopy.
Since it is 100% stable in animation and also camera independent, since the lines and surfaces are really three-dimensional (which isn't the case for most other NPR solutions), I can say that it looks fantastic in S3D!
The fine detail actually helps the eye solving the depth.

I'd really love to do a stereoscopic project in this kind of look :-)





Cheers!

Tom
Wow! Those look fantastic! I hope we can do this in modo soon. Is there any good tutorial on how to do this in Lightwave?
Catsmeat: Cool Avatar! ;-)

Check out Wacoms/Gideons Link in the first post.

But the basic principle is very simple:
You apply a procedural displacement to your object.
Depending on the look you want, you need more or less subdivisions (more = finer).
This creates the "jittery" look.
The stronger the displacement, the stronger the jitter.

In Lightwave you have to make sure you use a real 3D procedural, not a 2D one.
Fractal Noise, Turbulence and Crumple are the go-to choices here with increasing rendertime but also increasing interest.

Then you animate the position or offset of the procedural texture on the subframe level,
that means for instance that you move the procedural texture for a meter in X, Y and Z on frame 1
(the distance doesn't really matter but creates different looks (smaller = smoother, larger = more drastic)
so basically within one frame and then you set this animation (that you can't see in the viewport) to repeat,
so that it happens for every frame from the two keyframes you have set.
It is important to use repeat, so that the animation is exactly the same for each frame
which prevents flickering in animations.
This stability makes it so useful.

Should you actually WANT the Lines to move, I'd rather layer a second procedural on top and animate that over a longer frame range,
this gives a lot of control over both the movement and the look while doing both with one procedural
can be harder to control - but it depends a lot on the project.

Finally, for the magic to happen, you activate the old, "Classic" Motionblur,
the stepped pain in the behind from the dark ages which is so very very useful for NPR ;-)

"Blur Length" can be typically set to 100%.
The number of "Motion Blur Passes" define the look: in my above examples, the greyscale one uses something like 5 passes,
so the individual lines are more visible and create a rougher pencil look,
in the greenish-blue example the number of MB-passes is very high,
therefore you get a more chalky look.

As for the rest: you can combine the different Lightwave "Edges" as fits your taste
(experimentation is everything here)
and even cellshading - which I normally find to be a pretty dead CGish look and not like too much -
becomes very usable since the displacement roughens it up.
Both examples above use it.
But it can also be interesting to experiment with image textures.
Often pretty strong stuff yields interesting results,
like Black and White lines or cranked up wood textures etc.

Then you can dive into the details:
If you make the 1-frame-animation of your procedural not use a linear interpolation but a curve,
you can for instance have most of your lines crowd very close together and smooth,
and then 1 or 2 lines break away for a more crazy look.

Another thing is, if you want different amounts of jitter in different areas of your object,
like for instance in the face and at the fingers of a character:
This is very easy to do, if you layer a weight map on top of your 1-frame animated procedural
and paint the amount of jitter/displacement you want to see in each area.

For the crazy-text above, the plan was, to oscillate between the text going completely crazy and break up into chaos
and calmer moments where it's clearly readable and fluctuates
only slowly by a second procedural moving through the object.
This stuff is very controllable!

For the ultimate control, you could even sculpt different jittery objects in ZBrush or modo
and sub-frame-blend between the morphs. ;-)

While I explored a lot of techniques over the years, this still is the most simple, stable,
available and versatile that I know.
And in Lightwave it is blazing fast, because of the extremely fast edge rendering.

I hope some of these options will become available in modo in the future!

Feel free to ask if something is unclear.

Cheers,

Tom

Message edited by ThomasHelzle on 5/22/2012 - 2:28 PM

Quote from Gideon Klindt :


Sweet baby cheesus!* No wonder I couldn't get reliable results- this is enirely dependent on sub frames for non jittered line animation. Ok, need to keep my noob thinking hat on (and RTFM more) and keep plugging away- I had almost given up on this method in modo...will investigate- thank you Brad!

PS- what the heck are you doing on the message boards after the launch on a Sat night? You a) have a very understanding family b) a great work ethic c) must be more into 3D apps than most developers d) be obsessed with your own creation. I can only say thanks to all of those- and how amazing is it that the prez of lux is watching over his flock of users- just awesome- thanks!



*sweet baby cheesus is a registered trade mark of JavisJones Inc.


Pretty scary, isn't he : )
Thanks so much for the long explanation. I love the hand painted feel the images get. I think it would be very cool to try doing some animations with this technique. I want to try it on character animation. And I'm also curious about more techincal illustrations using this look. It might make them stand out a bit. Controlling areas of influence using weight maps sounds almost too easy. I'll follow your steps and try it out this weekend when I have the time.

Heh, and my avatar is the cinaminni monster (fast forward to 3:04), probably the scariest of all monsters, ever.
OMG - that is really scary LOL :-)

Keep us updated on your results - and especially if you should find ways to implement this in modo in interesting ways!

Cheers,

Tom
Page 1 of 2
   1  2  Next  Last