View Full Version : Martin's Technotes
08-09-2003, 11:08 PM
I just thought that I would draw attention to this:
Bits of it might be considered controversial by some of the people here but there is some interesting stuff.
08-09-2003, 11:36 PM
I havent read all of them but the ones Ive read dont seem controversial. Did I miss something?
08-10-2003, 01:10 AM
Before answering this, I would like to say that AM is an amazing character animation program.
I don't know whether contaversial is the right word, but it echos some of the earlier discussion on this forum. (Eg. Hash splines vs. Sub ds).
"A typical image rendered in Animation:Master is very fast, one of the fastest in the industry. "
Well I don't know all that much about the speed of other renderers but I am not sure about that comment. I for one am looking for a faster renderer for v.11 particularly where area lights are concerned.
I also don't like the idea of increasing the bias on raytraced shadows to get rid of terminater artifacts as it just brings on some of the effects of z-buffer lights (eg light rims under noses etc.). (Look at the bottom of this):
A lot of people are hoping for the modelling tools to be fleshed out to make things like mechanical bevelled objects faster to make but here he is suggesting that it won't happen. Well, to further his analogy, my attempts at making machines out of hash patches result in machines that look like they are made of clay. Maybe I have to practice some more but it doesn't take much practice to do the same thing in a sub d modeller.
Again, AM is a great program.
08-10-2003, 03:39 AM
Originally posted by John Keates
"A typical image rendered in Animation:Master is very fast, one of the fastest in the industry. "
I love AM and the new renderer is a huge improvement over earlier version....but.... it is quite a bit slower then current generation renderers. IMHO, a lot slower. And boy howdy if they don't keep an eye on what is coming and start really paying attention to current SIGGRAPH presentations, they will fall even further behind when next gen renderers come out.
Still, the boys at Hash deserve a pat on the back for what they have achieved in 10.5! :thumbsup:
08-10-2003, 04:59 AM
I kind of get what he was saying about sub div vs patches. Since sub div are using polys to make smooth surfaces. You need a bunch of polys to get a smooth look. But I agree that they do need to add more tools to help with mechanical modeling. I think it takes to long to model some stuff. ANd I agree that the renders in 10.5 are the best that Hash has ever been -------- but!! Damn is it slow!!!!! They need to speed this up-alot!
08-10-2003, 08:29 AM
You know... this comes around every so often and let me just say what I always say: Quality first.. speed a very distant second.
Ask the old timers what the best render was (pre 10.5) and odds are you will hear 7.1 mentioned. ask what the slowest was and you will hear the same number more often than not. 8.0 sped things up a great deal and introduced quality compromises. the now infamous black specks among others.
10.5 has the best render since v4 IMHO lets not start tempting them to make quality sacrifices to silence the speed complaints.
just the old wizened voice of experience here...
08-10-2003, 09:08 AM
Cinema4d, better quality, much faster renderer
Mental Ray, better quality, faster renderer
Renderman, even better quality, much faster renderer
Saying that quality must be sacrificed for speed, when comparing AM's current renderer to what else is available, is wrong. No matter how wizened one may be. Besides I think Martin is the kind of guy that constantly needs to be challenged. 10.5 wouldn't be what it has turned out to be if it wasn't for the community stepping it up.
And I would agree, after 10.5, 7.1 was the best renderer.
08-10-2003, 10:55 AM
My Fault, that was a good point well made.
If Martin cannot make the renderer fast and good then it is time to make export to something that is fast and good possible and easy. There is an increasing wealth of free renderers out there and they are only going to get better.
As I see it, there is no choice. The renderer should get better and faster every six months. That is what everyone else has been doing for years. Standing still is not an option.
Look at it this way: The best way to get a quality renderer is to make it faster. That way you can have more samples for AA, motion blurr, DOF and soft shadows. With a slow renderer, these things are not practicable.
Granted, if there are render errors then fixing them takes priority but having to put up with render errors so that render speed is increased is not something that Cinema 4D users for one have to do.
Oh, and yes, v10.5 is looking lovelyer all the time.
08-10-2003, 06:02 PM
My friend who uses C4d v8 just finished a quick animation. A glider flying through the clouds and some logo stuff. I asked about the render speed. While he said that C4d is very fast-as soon as you start adding the cool stuff, like the way you do with AM-it got very slow. Not as slow as AM-but from talking to people who also use LW-the renders do get slow. While I dont think AM is blazingly fast:D Im glad the quality is better. But Id like the best of both worlds-the quality and some speed improvements!;)
08-10-2003, 08:24 PM
The tech note mentions this:
Notice, that this discussion does not try and say either method is superior – the issue is that using subdivision surfaces precludes the use of Animation:Master’s animation technology.
No mention of superiority here, just an admission that the current state of Hash's animation technology is incompatible with the use of subdivision surfaces.
A matter of semantics? It also mentions this:
In 3D computer graphics, there are three disciplines: modeling, rendering, and animating. No one program is the best at all three. Animation:Master is the best at animating.
I agree with the above quote as well. Most of my 3D experience is based mostly in Lightwave and Maya. I got to AM much later.
Modeling right down to veins bulging in an arm, or even folds of the skin on a finger is very much possible when you model in Lightwave's version of subdivision surfaces (yes it's not TRUE subdivision surfaces but more than good enough to get the job done).
Hash's spline paradigm works well for much simpler surfaces. Getting a smooth surface, never mind a complex one is very challenging. One could spend weeks on a model. Something that could take just a day in Lightwave and is much less complex.
I take issue with rendering. Lightwave has a beautiful renderer. While I have no experience in Cinema 4DXL and have no idea how fast it is, Lightwave's renderer beats down the Hash renderer without much trouble. Much better quality and very much faster as well. Hundreds of time faster. I'd say at least 400% faster.
.he said that C4d is very fast-as soon as you start adding the cool stuff, like the way you do with AM-it got very slow. Not as slow as AM-but from talking to people who also use LW-the renders do get slow.
This is true in Lightwave if you use volumetrics, area lights, radiosity, etc. It gets very slow. But you don't have to use radiosity, you can use lighting techniques to make it look as good and render much faster, ditto with volumetrics and area lights. It still looks much better than Hash's render and at the fraction of the time.
Using multi-pass makes the render better in AM but makes it a lot slower. Let me rephrase that, a LOOOT slower. I test rendered the Toy Story scene and it took a LOT longer. Somebody benchmark it? I think it took me 20 hours on a 2 gig Athlon XP using standard D1 resolution.
Better quality yes, but at what price? The multipass feature is useless for my purposes, using a small render farm, lots of composites and corrections in a paint program was our way to bypass that bottle neck.
We're using AM on a fast track project, the decision to use the program hinged mainly on it's amazing char-anim tools and the relative simplicity of the characters we get to work with, and of course the price. If we had the budget to start with any tool, we would have probably gotten Maya or Lightwave (plus Keytrak and a host of other free plugins).
But the Lightwave engine is probably the slower renderer when compared to Renderman. Much slower. But a helluva lot faster than AM.
If Hash can up the ante, get the renderer up to speed and improve the quality... Hell yeah, DO IT. The Lightwave engine is five years old, maybe older... So they won't change the Hash patch paradigm, the animation tools don't need anymore work, just work on the renderer eh?
08-10-2003, 10:26 PM
perhaps for v11.0 we have to fight one more time.
As sayd before v10.5 is a nice version because of help of the community. We have to search the most important features and everyone has to mail hash and please them. Perhaps this works?
08-10-2003, 10:48 PM
Originally posted by koon69
I kind of get what he was saying about sub div vs patches. Since sub div are using polys to make smooth surfaces. You need a bunch of polys to get a smooth look.
Er... Not really. At the user level, your low-poly sds cage is just as light (in some cases lighter) as a similar spline cage. The advantange of sds, however, is that you can easily add localized detail by cutting/extruding/etc. polys. This is much more difficult (sometimes nearly impossible) with spline cages. The simple fact of the matter is that sds are better than patches in the vast majority of cases. Of coarse, if you're "used to" splines, it's going to take some time before you start seeing the benefits of sds.
EDIT: But, yeah, A:M still kicks ass in the animation department. That's why I upgraded to 10.5;)
01-15-2006, 09:00 PM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.
vBulletin v3.0.5, Copyright ©2000-2013, Jelsoft Enterprises Ltd.