如何假造 final gathering 的效果

[3dartist]

光線與render對3D來說是最後的一環，也是二個很重要的元件，目前好多的3D軟體都使大家認為，3D就是要從制作到最後的render，都要一次完成，其實這是一個很嚴重的錯誤觀念，因為3D軟體公司，為了銷售軟體，它們必須在3D軟體中一次完成，但是在業界這是不可能的，一來全在3D軟體中，完成無法達到因有的水准，或是就算得到了標准，可是太耗費時間，所以一般在業界都分層render，再到composting軟體合成，這樣不但節省時間，而且如果發現有錯誤，或需要更改時只需要，render該物件在合成即可，當需要更改物件顏色或調子時也只需到composting軟體修改即可，這可以省下大量的時間，也可以做出較好的效果
<a href="http://userwww.sfsu.edu/~cwchang/3dschool/movies/BALL_BUNCE.avi"><img src="http://userwww.sfsu.edu/~cwchang/3dschool/images/final_s.jpg"></a>
<a href="http://userwww.sfsu.edu/~cwchang/3dschool/movies/BALL_BUNCE.avi">DIVX Movie 200kb</a>
以下連結是制作方法
<a href="http://userwww.sfsu.edu/~cwchang/3dschool/lightcompos.htm">如何假造 final gathering 的效果</a>

<a href="http://www.c-c-w.com"><img src="http://www.c-c-w.com/banner.jpg">www.c-c-w.com</a>

[Wolf]

非常好的一個作業哦。很希望以後能看到多些這類教學帖子。

用後期合成來控制畫面的確是行內standard的做法。不過也有不用這做法的成巨狺l，如Pixar在Monsters Inc很多場景試驗用“一次過”的渲染方法，盡量不用comp，效果也相當棒。

但是我倒有另一種想法：近年有幾個3D軟件靜悄悄的在開發新的渲染概念，特別是即將在下星期發表的Houdini5，裡頭有個叫deep raster image的東西。用它加上適當的合成技術可以把所有渲染設計工作推到最後階段去。受影響的工作范圍包括texturing，lighting，甚至有限的staging。還有Pixar之父Ed Catmull傳聞正在研究沒有pixel的圖像... 其實現在Prman已經可以渲染arbitrary data圖像，只要有適當的合成工具，就可以全程在comp裡頭做render。渲染顏色圖的時代將近尾聲了。

[Bigeye]

Could you explain more about the "deep raster image" and the concept of "no pixel" This topic is quiet interesting.

Thanks : )

[Wolf]

Quite simply, deep raster image is a 3D image (almost all images that we've been working with are 2D), so you can store your 3D scene inside it. You can "walk" into it, thus the name *deep* raster image. You can use it to visualize cloud, water, blood veins, math functions... and most interestingly you can use it to do 3D (not just 2D) compositing. You have to see it in action. It's incredible!

As to pixel-less images, well I don't really have any information. There's a website that talks briefly about it in an interview with Ed Catmull, I'll put up a link when I find it.

[3dartist]

<a href="http://www.softimage.com/Products/Xsi/v2/video.htm">http://www.softimage.com/Products/Xsi/v2/video.htm</a>
in Softimage|XSI 2.0 have the composting funtion. Is this Wolf talking about?

[Wolf]

No. The incorporated compositor that XSI boasts about has been available in Houdini for over six years now. Deep raster is a different thing altogether - it is an extension of the idea of "image" to a wider function of 3D data repository. Think voxels. Digital Domain recently used a similar image object that they developed in-house to create the water horses in LOTR.

[Bigeye]

It sounds like the particle simulate in the 3D data or 2D image(It seems like the particle attach in bump mapping). Moreover, I think the particle (indepentant) could changing their feature to suit the enviroment. May I ask the question ? It's that means the particle occurs in the compositing.