This document discusses how to maximize production efficiency through efficient content authoring tools and pipelines for inter-studio asset development. It covers topics like infrastructure, content authoring, data conversion, game runtime considerations, and asset sharing. The key recommendations are to identify and reduce latency throughout the production pipeline, automate processes as much as possible, prioritize tool and workflow improvements, and treat outsourcing partners like internal team members to maximize output.
TrustArc Webinar - Unlock the Power of AI-Driven Data Discovery
Maximize Your Production Effort (English)
1.
2. Maximize Your Production Effort
“Supersize your Production Pipe!”
Efficient Content Authoring Tools and Pipeline
For Inter-studio Asset Development
3. Introduction
• Who am I?
• Who are Slant Six Games?
• Who are you?
4. Paul Simon Martin
• Director of Technology, Slant Six Games
• 14 years videogame industry experience
• Slant Six Games („05-Present):
• Technical Direction: BigBrain Game Engine
• AAA-quality, cross-platform engine
• Xbox 360, PS3 & Windows
• Full suite of content authoring tools
• Data transformation pipelines
• Automated build, test & publish environments
• Sony Computer Entertainment America („00-‟05):
• Developed: PS2 low-level & graphics engine (Syphon Filter)
• Developed: PSP low-level & graphics prototype engine (Syphon
Filter)
• Prior („96-‟00):
• Graphics engine & tools for PS1, Dreamcast & Windows
5. Slant Six Games
• Founded in 2005
• Products:
• Unannounced Title – PS3/Xbox 360/Windows 2010
• SOCOM U.S. Navy Seals Fireteam Bravo 3 – PSP 2010
• SOCOM Confrontation – PS3 2008 and Cold Front DLC 2009
• SOCOM U.S. Navy Seals Tactical Strike – PSP 2007
• Syphon Filter: Dark Mirror - PSP 2006 – Rendering Engine
• Awards:
• BC Business Magazine - Best of BC Top 10 Companies to work for in British Columbia - 2009
• Elan Award – Best New Game Company – 2008
• SOCOM US Navy Seals: Tactical Strike
• Elan Nominated - Best Hand-Held Game of the Year - 2008
• BAFTA Nominated - Best Strategy Game - 2009
• E3 Best PSP Multiplayer Experience -2007
• Technology
• AAA class multi-platform (PC, XBOX 360, PS3) Proprietary Game Engine
• Proprietary Rendering and Animation Engines , Pipelines and Toolsets
• PSP proprietary Engine and Toolsets
6. Outsourcing @ Slant Six Games
• Started outsourcing art content in 2007
• Locally only at first…
• Then China
• Currently working with
• 2+ partners in China
• 1 partner in Canada
8. Glossary of Terms
• Source Content:
• Any unoptimized game asset file that has not yet gone through
data conversion
• E.g. *.ma, *.dds, *.tga, *.csv
• Game Data:
• Data file that is ready to be loaded by the game
• Usually optimized for platform(s)
• Authoring Tools:
• A tool (usually residing on PC) used to author source content
• E.g. Maya, Photoshop, CustomStudioTool
• Data Pipeline:
• A set of tools & scripts that are used to convert Source Content
Game Data
• Typically driven via a build engine
9. Glossary of Icons
• Slides
• Online content
(will discuss during this talk)
• Offline content
(either hidden slide or will gloss over)
• Suggestions
• Practices that work well for us
(TODO: xxx)
• Things we learned the hard way
(TOAVOID: xxx)
13. The Evolution of Production
Circa 1994 AD
• Part 3: Somebody hired a Game Designer
14. The Evolution of Production
Circa 1998 AD
• Part 3.5: Artist & Designer Spanking –
popularity diminished
15. The Evolution of Production
Circa 2000 AD
• Part 4: Can‟t we all just get along?
16. The Evolution of Production
Current Day
• Studios value their artists and designers
• Artists and designers:
• Build The Game
• Programmers:
• Support artists and designers
• Implement core gameplay that can‟t be
data-driven
17. “Your Studio is as good
as your Tools”
• Paul M, circa 2003
• “Tools”- Authoring, Infrastructure, Pipeline, Packaging
• Today‟s games must have massively efficient runtime
=> Bake as much optimization into data as possible
• Games built by production teams
• Faster workflow => more content
• Faster workflow => more iteration = polish = quality
20. Supersize your Production Pipe
• Production pipe (prə duk′s̸hən, prō- pÄ«p):
• “The process that generates content for your games”
• narrow constrictive poor flow
• wide dilative good flow
• Considerations:
• Internal team process / workflow
• External team process / workflow
• Authoring tools / workflow
21. Supersize your Production Pipe
• Super-size (so̵̅o̅′pər sīz′):
• “Increase capacity”
• “Make bigger”
• In order to:
• Increase flow
23. Latency – The Silent Killer
• Latency: A measure of time delay
experienced
in a system
• High latency = Low flow
• Identify maximum iteration time
• Identify latency
• Derive a plan
• Re-evaluate
27. Old School
1. Trigger volume authored/tweaked
in the level editor
2. User hits the “export” button
3. User runs a script to re-export the
entire world
15 mins
• textures converted, visibility map
compiled, navigation mesh
generated, …
4. User waits 10+ mins
5. User launches the game
6. It takes 5 mins to load
7. User runs player character from
level start -> test area
8. Volume was too big/too small? Try
again, back to step 1.
28. Much Better
1. Trigger volume authored/tweaked
in the level editor
2. User hits the “save” button
3. The build infrastructure detects a
file change
5-15 secs
• Triggers fast asset conversion for the
new trigger volume alone
• Signals (already running) game
there is a change
• Game hot-loads new volume and
associated scripts
• Player character (still) in test
location
4. Volume too big/too small? Back to
step 1.
29. Author vs Tweak
• Author:
• Addition/deletion of something new
• Latency: 1-30 seconds
• Tweak:
• Movement or property change of
something existing
• Latency: < 5 seconds
30. What Happens Next…
Post-iteration, user commits changes…
1. User hits “commit” button User involvement
ends here / moves on
2. New level data committed to to new task
asset depot
3. Build/Test Server (BTS) detects a change
and begins asset conversion
4. BTS launches an automated test of the game
with new asset
5. All happy? BTS sends user a confirmation
email
6. Problem? BTS sends user an email with
relevant info
38. Infrastructure - Sharing Tools
• Which to share and why?
• Remember that dotted „sharing‟ line a few
slides ago…
• Workflow for sharing resultant data assets
• Tool & data versioning
• Developer studio tool versions usually more
recent than remote studio‟s. So…
• Flexible runtime or…
• Asset upgrade process
42. Content Authoring
• Intuitive UI
• Also, mirror existing commercial tools key
bindings & shortcuts (Maya, Photoshop, …)
• Consistent UI
• Close to Zero-iteration time
• The 5 Second Rule
• Ideally in-game/in-viewer
• Or in tool viewport
43. Organizing Assets
• By meta-tag (e.g. iTunes)
• Requires implementation effort
• Tends towards poor file management
• By directory (e.g. Windows)
• Tough to relocate
• Find a balance
• <= 200 tagged items per folder
• Disallow tags defined by user
44. Tracking Assets
• Tracking software/tools
• Hansoft
• Filemaker Pro
• Excel
• Consistent language
• Normal map change
• Mesh change
• Detail change…
45. Budgets and Metrics
• Establish „Cost Scores‟ for all assets
• Find a currency for each asset type
• E.g.
• Filesize (Kilobytes)
• Dimensions (pixels wide/pixels tall/pixels deep)
• Runtime cost (GPU cycles or even # lines of code!)
• Allow users to evaluate the total cost of a
scene
• Document currencies for outsource
partners
• Document budgets in asset tracker when
outsourcing
49. Data Conversion
• DOS Batch files are evil!
• Limited functionality
• Poor scripting language
• Slow to execute
• No data dependency checking =>
each build takes as long as the
previous
• Hard to debug
50. Data Conversion
• Many better options available:
• Scons (we like this )
• MSBuild
• Ant/NAnt
• Jam
• Make
51. Data Conversion
• Error Checking:
• As much as possible
• Report in a timely manner
• Often convenient to check for missing assets in
data conversion pipe but…
• Try to prevent errors getting this far
• Better to report errors at authoring time
• Report errors clearly!
53. Engine Considerations
• Middleware:
• Not inherently evil but don‟t rush into it
• Consider: Tools authoring process (identify any latency!)
• Consider: Interoperability with your runtime
• Consider: Integration with your tools & data
pipeline
• Consider: Support?
• Consider: Source code?
62. Asset Sharing – Package Drops
• Problem:
• Long build process for packages
• Usually commence late in the work day
• Someone needs to work late!
• Solution:
• Automate the package shipping!
64. Don’t
• Let production process evolve on its own
• Always prioritize game features over tools
& workflow improvements
• Ignore your production team‟s workflow
requests
• Underestimate management effort for
outsourcing
65. Do
• Identify latency in your production pipeline
• Automate as much as possible
• Have regular „asset authoring priorities‟
meetings
• Hire an army of tools engineers!
• Create an outsourcing plan
• Treat outsourcing partners like part of the
team
• Regular status meetings with outsourcing
partners
• MAXIMIZE YOUR PRODUCTION PIPE!
Programmers sat in dark offices and built an entire video game.Some of them that had a partial right-brain, also created the artwork
The chance of hiring a programmer that has good artistic skills is slim-to-nonePeople started to realize this fairly early on and hired a token artist
Oct 2008: Richard Garriott is the first game designer in space.The creator of the Ultima series of video games was launched into orbit with a crew aboard a Soyuz TMA spacecraft.
Some assertions…Quality comes from iteration – more iteration cycles, higher qualityMore time = more iteration cycles = more quality
In the land of the AAA product, Iteration is king!
Because you…
What breaks flow?If you’re as impatient as me, then you are acutely aware of latency in almost every consumer electronic device that you interact with, daily.For example, drawing cash from an ATM, authorization after punching your PIN into a supermarket checkout, etc.In a production environment, the time that is taken by artists or designers waiting for results to happen after they hit the “do stuff now” button is wasted time that is hard to hide with multi-tasking.What I mean by that is that typically they are either:1/ Too short to switch gears and do meaningful work while waiting for the results OR2/ Too long to wait for the results, so the gear switching is a long one and the person loses momentum on that task altogether
The rest of this talk will be divided between 6 sections related to the Production Pipeline
To illustrate my breakdown of “the pipeline” here is a slightly contrived example of the lifecycle of a typical new game feature, in an authoring pipeline... PC enters instance of EvilThingSpawnTriggerVolume, causes evil NPCs to be spawned that want to cause PC harm
Some rules of thumb…
Take a step back, look at your process Take inventory, see what works well, what is broken Can you draw a line that easily separates the tools / data that you wish to share with an outsource partner?
Here is a dump of some execution timings for tools in our Data Conversion pipeline Can you generate something similar for analysis?
Pre-define meta-tags (e.g. Building, Detail, Blocker, Light, Trigger) – don’t allow each user to define new ones that are shared by the team Document your pre-defined tags for consistency with outsourcing partners
Pre-define meta-tags (e.g. Building, Detail, Blocker, Light, Trigger) – don’t allow each user to define new ones that are shared by the team Document your pre-defined tags for consistency with outsourcing partners
Wherever possible, source assets should have a clearly defined associated cost.Establish a currency that is well-defined (even if using a meta-points system) and used when referring to budgetsWhen working with an outsourcing studio, this system is invaluable since it allows a black-and-white way to define your expectations regarding asset budgets, and whether they have been met.
Asset meta-currency Each coloured dot = different resource type
Scene cost barometer
DOS Batch files are evil!Seriously, if your studio’s data conversion pipeline consists solely of a bunch of batch files that are run sequentially, take it out the back alley, shoot it, and tell your kids that Floppy died a peaceful death and we’ll get a new one.
I have presented entire lectures on this topic so please free to ping me offline if you have any questions here
Runtime interoperabilityMemory managementProfilingTools IntegrationDo they supply format conversion tools?Can you drive their data conversion tools from the command line?Do they expose an API for deriving input dependency files and resultant output files?
These are non-trivial features for your engine to support Can’t have engine team working in isolation of tools team
Generation of TRC-compliant PSN / XBL packages is a complicated process Familiarize yourself with it early One of the best decisions I’ve made was to set up our PSN package authoring process really early in development – it exposed a dialog with Sony’s external TRC staff that would have delayed the shipping of our game if we hadn’t addressed it then. Don’t leave until Beta!
Package drop Package up assets on each end, send to recipient by some protocolFTP – slooow (but free)Aspera – fast (but costs) I have good personal experience with Aspera; it was our transfer protocol of choice when working with Sony Japan and Sony USA
Problem: Someone needs to hang around after work hours, waiting on a lengthy build process, just so they can push a button to start the FTP/Aspera transferSolution:Have your package build pipeline automatically post the package upon successful completionEasy to do using FTP command line clientI worked with Aspera’s support people to customize their command line client such that we could automate delivery by FASP – now also trivial to implement
At least 1 manager fulltime to manage outsourced assets