Datasheet

38 chapter 1 ■ Setting Up a Composite
Nuke 5.1 logo
LL: Nuke was developed in-house at Digital Domain. Describe the path it took from
Hollywood to The Foundry.
RS: Nuke was [written] as a command-line image-processing program.… [In 1993,] it was
used by Digital Domain on the film True Lies. The first version, [which] became Nuke 2,
was created to read scripts, draw, and show data flow down the tree. It was used for Titanic
and the T2 3D ride along with many other films.… Nuke 4 was the first commercially avail-
able version…from the newly-formed D2 Software in 2002.… Selling commercial software
is outside Digital Domain’s core interests. Both Digital Domain and The Foundry real-
ized that Nuke could have far greater potential if it was developed and marketed by The
Foundry. Today, The Foundry owns the Nuke IP (intellectual property).… Digital Domain
remains one of The Foundry’s most important customers.
LL: Nuke is node based. What is the greatest advantage of a node-based system over a layer-
based system? Are there times when a node-based system is a disadvantage?
RS: A node-based system allows you to change earlier operations without having to repeat
the later steps manually. It also allows a free flow of operations, whereas a layer-based sys-
tem will direct an artist’s approach to working with a particular application model. Such
freedom in constructing the flow of data and operations promotes many advantages, such
as intermediate result reuse, the creation of operation templates, more creative and powerful
use of scripting and expressions, 3D data flows, and nonimage data use.… A disadvantage is
that the program actually does allow such flexibility and may repeat the later steps for you,
so this can be slower when you change a step other than the last one. A layer-based system
can support a better time-based view, useful when working with multiple shots together in a
sequence.
LL: Nuke has always operated in 32-bit floating point. How would you summarize the
advantages of floating point over lower, fixed bit depths?
RS: Even in 1993, floating-point calculations were faster than integers on modern proces-
sors. Floating point also offers a wider latitude, [which] allows linear data representation
rather than having gamma to encode detail in specific areas.… This allows artistic and tech-
nical freedom during image processing, [thereby] removing the need to carefully manage
clipping in the data. For example, this is useful in color correction (where values are pushed
up and down multiple times in succession), with depth or motion data (where great accuracy
is needed), and in HDR workflows that need to support a wide range of exposure values.
52615c01.indd 38 10/20/09 12:34:26 PM