|
 Features
Keeping
Up with the Sims:
Managing Large Scale Game Content Production
With project
budgets in the multiple millions of dollars and virtually no margin for
error, more and more development teams are under tremendous pressure to
come out on top of the entertainment software market's cutthroat competition.
No team manager wants to contemplate dropping the ball when creating the
vivid graphics necessary to help make a game a success. Electronic Arts'
The Sims franchise is an excellent example of this pressurized
situation. With recordbreaking sales on its initial release of The
Sims, as well as the subsequent successes of the expansion packs The
Sims: Livin' Large and The Sims: House Party,
EA had a lot riding on the success of its content generation strategies.
Much of The Sims' appeal is due to the very same element of the
game that presents one of EA's strategic production challenges: the sheer
number of highly detailed, technically complex objects that populate its
game world and provide interaction for The Sims, the little creatures
that are at the core of this unique world. Designing, specifying, engineering,
and troubleshooting these objects is no small amount of work; EA counts
on balancing that effort by send ing the actual production of much of
the sprite art assets for these objects out of house. New Pencil has been
the studio charged with delivering on EA's demanding vision for these
critical assets, from providing auxiliary production capacity on the original
version of The Sims to the creation of most of the sprite assets
on later expansion packs, such as the recently released The Sims:
Hot Date.
This article
highlights the critical issues that govern the highvolume asset production
needed for today's most demanding games and some of the techniques upon
which New Pencil has relied to create the artwork for The Sims
franchise. These principles were developed at New Pencil and at Maxis
over the long association between the two studios, and New Pencil has
found these techniques to be fundamentally good practices. Although New
Pencil uses these principles to provide highvolume content outsourcing
services, they can be employed by any internal development team with excellent
results. If you are in the position of managing this kind of production
capacity, there are some things you are going to need to consider.
As New Pencil
developed its systems over repeated contracts with Maxis, four strategic
targets emerged that helped us generate a large body of game content on
time and on budget while still holding quality as the uppermost value:
time estimation and cost modeling, style matching and crossteam consistency,
asset staging, and project tracking and team management.
These four
fundamentals are cornerstones of a solid production foundation. Omit one,
and your production may collapse. Let's take a closer look at what each
target entails.
Time Estimation
and Cost Modeling
No project
is going to be delivered on time and looking great without good projections
of how long each asset will take to make and how much it will cost to
make it. Too often, teams stop short and focus exclusively on the underlying
technical issues that support time estimation without going to the next
step. This is especially true when work is being done internally, where
budgets often deal with large numbers of assets as a group line item such
as "animations."
The reason for doing these projections goes beyond just wanting to hit
a budget and a timeline; it goes to the deeper issue of quality. Good
time projections allow you to reserve buffer time for extra polish, integrate
new techniques, and create tailored resource plans to handle special concerns
in the production without having to rush too fast or cut too many corners
on the look of any asset.
Once the time estimates are in place, it is then possible to institute
good cost models. A solid cost model allows internal teams to make the
best use of their budgets and outsourcers to defend their profit margins,
which in turn allows them to be more flexible and serviceoriented toward
their clients. It seems easy just to take the average time you think an
asset will require and multiply that by the number of assets. At that
point, you need to then add 20 percent right away as a rule, as a rough
stab at accounting for the unknown developments that are sure to occur.
No seasoned client or production manager, however, wants to depend on
that kind of superficial analysis; he or she knows that there are far
too many efficiencies to be found and dependencies to be contended with
for any such simple formula to reflect reality. In order to develop a
more indepth cost analysis, we focus on the following three general areas:
the pipeline, the approval process, and the aesthetic target.
The place to start in developing good time estimates is the pipeline.
Time estimates need to be based on a pretty solid production pipeline,
or else they are meaningless. This is not to say that a pipeline can't
be upgraded or amended during production, but unless the baseline is well
understood by everyone who is going to use it, those upgrades cannot be
evaluated for their impact on the schedule. Begin by posing a few questions
about the pipeline: Who are the experts in the use of the pipeline, and
is there access to those individuals? If assets have already been through
production on this pipeline in the past, what were the timeframes associated
with them? What technical taboos or requirements are there in the creation
of the assets, and are they documented? How modular is the pipeline? If
an error has been made somewhere during production, how far back in the
creation must the artist regress to bring the asset up to specification?
We relied successfully on Maxis to provide instruction and support on
their homegrown export tools and worked to make sure the pipeline that
we set up at New Pencil was as close as possible in implementation to
the pipeline at Maxis. The Sims ' sprite pipeline was well developed
before Maxis went casting for a contract house to expand their capabilities.
Just as importantly, Maxis worked hard to ensure that New Pencil had access
to its experts who had the inside track on how the pipeline really worked.
Maxis had clear expectations as to how modifications were to be made to
assets and at which point in the pipeline each asset should be modified.
New Pencil also had a team with a long history of working together, which
is a huge plus when developing time estimates. Many new studios or teams
are forced to rely on experiences with other teams that may not reflect
their actual present skill sets, exposing them to potentially large errors
in time estimation. As a result, New Pencil had a high level of confidence
in estimating the time the average basic asset would require, excluding
revisions.
Understanding the approval process and the approval team is the next most
important factor. Are the members of that group in good communication
with each other? Are the list of reviewers involved in regular critiques
of the submissions restricted to only those with clear responsibility
for the final result? Is the review group casting a wellaimed net to garner
feedback from other individuals with influence on the product's future,
such as marketing or the publisher? Is there someone on the review team
whose task is to manage the review team to timely consensus and with the
freedom to make a command decision when the team is at loggerheads? Is
there a clear progression of stages that an asset will pass through to
avoid changes to previously approved work? Can the art team expect to
receive feedback from the reviewers in a timely manner? Is there a general
aesthetic vocabulary in use by the reviewers, and is the art team conversant
with it? In the case of The Sims, we were able to answer yes to
all of the preceding questions about the review process due to the amount
of preproduction work that Maxis had done, its own existing production
pipeline, and the closeknit nature of its team leadership. If the answer
that you get to more than one of these questions about your approval process
is "no," or if the questions cannot be answered at all, get ready
for rising costs. Sadly, it's easier to say that the costs will rise because
of a faulty approval process than it is to say by how much, but a good
rule of thumb is to multiply the costs associated with any existing approval
process by 1.5 when there is risk of multiple iterations.
For internal development teams, calling attention to this cost inflation,
if done diplomatically, can be a useful way of generating the leverage
at higher levels of the company to address the problems within the team
that underlie the approval process's instability. No executive wants to
see money being thrown away because there are questions about how solid
the review team is. For outsourcers, this is a more dicey proposition.
Ideally, if your relationship with your client is solid, you can consider
a gently candid conversation about your concerns.
If the relationship is not that strong, you're left with few options besides
simply charging more for the work in order to reduce your exposure to
these higher costs. However, in the bargain, you're also throwing away
your competitive edge. In any case, make sure the contracts that you sign
clearly state time limits and the format of feedback so that you can protect
yourself. It's no substitute for being able to help the partnership address
reviewrelated challenges, but it may help you avoid eating too many revisions.
Having a clear progression of stages for the assets was also crucial.
It helped focus the feedback from Maxis on the aspects of the asset at
issue and made sure that the likelihood of large revisions was greater
in the early part of development, when changes were less expensive. The
process that resulted worked to achieve ingame integration as early as
possible, giving Maxis lead time to support New Pencil should the assets
not perform as anticipated.
The process was divided into the following stages:
Design
evaluation.The asset is shown as a highresolution render in 3DS Max,
with no texture and only enough model detail to communicate the general
direction of work (Figure 1). The key elements of the asset as described
in the asset specification document are referred to, but may not be completely
modeled.
First model. The first model is shown as an ingame screenshot with
little or no texture (Figure 2). Color may be used to help differentiate
the model's substructure but is not yet a subject for review. All comments
from the design evaluation feedback are addressed, if they are relevant
to this stage. In some cases, the asset may be well developed enough to
skip the next step and go on to texturing.
Final model. The final model is shown as another ingame screenshot,
where the model is complete with respect to geometry (Figure 3). All sprite
animation states are represented, and the animations function in the game.
All comments from the first model submission are addressed, if they are
relevant to this stage. In some cases, there will still be model comments,
but they will be small enough to be addressed in the forthcoming first
texture stage.
First texture. Materials and lighting are shown for the first time
at this stage (Figure 4). All comments from the final model stage are
addressed.
Final candidate. Ideally, all comments have been addressed and
the asset is ready to be delivered (Figure 5).
The last
concern for time estimation and cost modeling is the aesthetic that you're
trying to match. It's safe to say that this is a larger concern for outsourcers
than for internal development teams, due to the separation between the
conceptual artist or art director who has the vision that sets the bar
and the art team executing the work. Nonetheless, the more elaborate and
esoteric the style is that has been chosen for the work, the more it will
cost to pro duce. Don't confuse this issue with the technical complexity
dealt with in the preceding pipeline discussion. Rather, this is the cost
of having to undertake the revisions necessary while the artists internalize
the aesthetic they must match.
The questions that need to be posed regarding the style matching have
more to do with one's own capabilities than what one needs from the client
or the art director: Do you feel that you understand this aesthetic? Are
the reference materials both broad enough to support a consistent look
across all assets and deep enough to communicate clear individual asset
identities?
Are the various skill sets that the project's aesthetic demands present
in a balanced fashion across the art team? Is the requisite familiarity
with software present in enough art team members to support general productivity?
Is there someone on the team who has a particular understanding of the
reviewer's vision who can share this insight with the art team? Again,
we were able to answer yes to all of these questions; this was essential
to New Pencil's confidence that the final assets could be produced with
roughly the number of submissions planned.
So let's say you've asked all these questions of yourself and of your
client or development team and you've come up with some time estimates
and costs that you feel pretty sure about. Next, you should gauge your
assumptions with a test run. Don't just sign up for the whole shebang
if you can help it. There are going to be hidden costs, surprise developments,
and plain old reversals of fortune you may or may not have anticipated.
Choose a representative set of assets a few easy ones, a few medium
ones, not too many hard ones and set yourself a milestone deadline and
a team structure that reflects your best guess as to time and costs. It's
important that everyone, especially your client or management, know that
this is a test, and that the estimates will likely change as a result
of how the test works out. It's a good way to blow the bugs out of the
system and get everyone on both sides of the art fence accustomed to working
together, without committing the entire project to deadlines and cost
expectations that just aren't realistic.
Early on, New Pencil and Maxis took the time to learn to work together
by beginning with more modest goals and then evaluating the realities
of production that emerged. As a result of these early lessons, the negotiated
price for content did rise by some 30 percent, but because the standard
of quality had been set and the relationship forged in a climate of good
communication, these price increases were perceived as reasonable and
did not have to be repeated later, which would have endangered the longterm
relationship between the studios.
|
|
|
|
|
Time
Estimation and Cost Modeling
