Re: XSIMAN

[Raffaele Fragapane <jaco(at)thejaco.com>]

Alright, to bring this back on topic...

Reliability of XSIman et similia:
You might need to detail the point further.
Why do you want to use renderman? Is it a project constraint (client 
wants RIBs), or is it a choice that will go from text files to pixels 
all in-house?
What renderman engine do you plan on using?
What is letting you down in MRay that makes you feel you'll need to jump 
through the hoops a new solution will obviously require?
What kind of work do you do?

Renderman is just a standard and a committee. The advantages of 
PRMan/3Delight/whatever over MRay are related to more then just the 
standards. Advantages (and disadvantages) are twofold, they come from 
REYES, and its architectural qualities, in terms of performance and 
scalability, and they come from RMan, in terms of implementation and 
deployment.

Without knowing what you want to do there's little advice that can be 
given that would make sense.


Affogato:
Affogato is always in development.
It's currently an in-house tool that evolves based on what projects we 
handle, and it soon will be an OSS that will need to face more then our 
in-house exigencies. So the answer to the first part of the question is 
very likely to be yes, it is still in development, and it will be for as 
long as Moritz, or anybody else who will contribute, feels like he has 
the time to maintain it.
The ETA of the release, what will be in it, and the other details I 
can't share yet, and that's not because we're sworn to secrecy or 
anything like that, it's simply because these are not easy things to 
clarify enough at this stage. It shouldn't be much longer is probably 
the only answer I can give.

What I can tell tho, is that it's more then barebones.
We've used it for every possible primitive, from polys to implicits 
going through nurbs and curves.
It Supports a lot of features for most of them (creases for subDs etc.), 
and it even supports some things that ootb XSI doesn't expose to MRay 
(like rendering curves).
It's very elegant and simple in the way it handles shaders, at least in 
the context of what is "RMan style", and while it has a definite bias 
toward the TD side, it only requires a lil bit of commitment to learning 
to be used, and not a 24/7 team of rendering gurus.
It's definitely stable and performing in terms of data throughput, as I 
still have to see something that will output the same sheer volume of 
data in any less then double the time affogato takes to do anything 
(some of the internal methods that had to be implemented to work around 
XSI's "structural choices" are just as fast, if not faster, then some of 
XSI's own API tools, which is one of the thing that irritated Moritz the 
most, as it basically pointed at some inner workings of the API being 
very inefficient, like several crawlers and retrieval operations being 
clearly as slow as parsing non hashed data and doing string comparisons).

It's not perfect probably, and it has to deal with some things in XSI 
that are very unfriendly to the RMan specs, but all in all it does a 
very good job of not making that a burden.


Rendering with renderman from XSI:
There's no recipe, because it hasn't been done very often.
To my knowledge the only two companies who've done anything relevant 
with XSI and RMan are the Mill, with XtoR and several top of the pile 
(no PS2 ad pun intended) commercials, and us, with Affogato.
There's a third company that might have done something, going by "what a 
friend told me", and there might be more I'm not aware of, but that 
would be just spinning the rumor mill and grinding some.


Last but not least, don't expect it to be an easy path at any time, and 
definitely don't expect a recipe to be found anywhere.
The public knowledge that renderman isn't friendly to small teams or to 
fast paced projects is pretty much BS, but we're not talking about VRay 
or some other EZmode toy engine. The things that make PRMan or 3Delight 
good products (and mental ray suffers from the same issue, altho in 
smaller measure since 3D software vendors had to bend backwards to fit 
it in) don't come for free, and they definitely have a price in terms of 
commitment and effort. All of those factors disqualify in the bat of an 
eyelash the idea of a "recipe".
If you want ootb and predictability of both strenghts and limitations, 
which is what I would consider "a recipe" Max + Brazil is probably a 
stronger set of tools.

******************************
|     Raffaele Fragapane       |
|     Rising Sun Pictures      |
| "Remember, TD is for TopDog" |
******************************



Prapas Akie wrote:

> They say a good discussion rewards promising results…
>
> Unfortunately, I didn’t see much feedback on my original question 
> about the reliability of XSIMan.
>
> Affogato has been added to this list, where does it stand? Is it still 
> in development? ETA?
>
> Is there a reliable recipe to get XSI scenes rendered with RenderMan? 
> Which recipe holds the highest probability of success?
>
> Thanks again in advance and have a good day.
>
> Akie Prapas
> 3D VFX
>
> ------------------------------------------------------------------------
>
> No virus found in this incoming message.
> Checked by AVG Free Edition.
> Version: 7.1.409 / Virus Database: 268.14.19/555 - Release Date: 11/27/2006
>
>  

---
Unsubscribe? Mail Majordomo(at)Softimage.COM with the following text in body:
unsubscribe xsi