So I was worries about the Warren Spector session being packed, so I decied to check out this audio session.  This session goes into the difficulty of managing a game with a great deal of v/o workflow.  The session has Greg deBeer from Sony Computer Entertainment, Ken Kato from Microsoft Game Studios, and Tom Hays and Rafel Lopez from a recording studio.

Sony has built a tool call LAMBS (could be wrong), which was developed by SCEE Cambridge.  It was built to allow for smooth audio localization on large scale projects.  Main focus of what Greg will talk about is script managment.  First draft of the tool worked with Word and Final Draft, and as well as using Excel because people seem to know it.  They are currently moving to basing the system to XML which would allow for a much more extensive oppertunties.  Once you have your script in Word or Excel, finding a way to build a class or bucket system that can be tagged in the XML output.  Such as mission 1 gets a tag of 'Mission 1' which would also allow you to search for very detailed information and build out readings or what has and has not been recorded yet.  As well as the XML is the naming convention, so something like a character named Jack who has a line in mission 1, would be JA_mission1_<idenfiier>.***, this allows for easy find as well as cuts down on duplicates.  Version control is a large pain, and one thing is to move the audio outside of the game tree in its own depot, and audio is just passed over in larger chunks.  This is still being figured out.  One thing that they have tried is using Text to Speech to place temporary audio in the game for timing and just overall fell and have the ablity to track these assets and replace them.

Ken Kato starts talking about the team they have developed interally within MGS Publishing called Gamma.  Gamma is a web based database system, the assets are protected by two firewalls that are 128bit encryption for security purposes.  Scripts are built in Excel and imported into Gamma.  The internet connection is really slow, so the demo is kinda falling apart.  The script finds a character and you can click on that character and it will list all the lines.  Metadata on the each line is stored in a database and things such as who someone is talking to and the tone they need to speak in.  Once audio is recorded, the line will become a link and you can just click on it to hear it.  Files are only named with numbers, which is generated from the line in the script.  This cuts down on the possiblities of human error.  Outside of the filename, it has a subname for outsakes, so 0001_out1.*** would be an outake for line 0001, where 0001 is the current approved version of the audio.

Tom is showing off his way of managing lines.  He has built around Excel and using batch file magic for renaming.  Using Excel and placing DOS commands within the actual fields and then export and running the newly created batch file to create file structures for locations.  It seems to work, but fairly convoluted.