Monday, March 16, 2015

My 2nd GDC - Still has that New Conference Smell

It was a year ago this month that I attended my first GDC, and it was at that time that I saw what would eventually become the Oculus DK2.  When I saw positionally tracked VR,  I knew that my new fun hobby was going to turn into much more.  It's been a ride of a year since.

So understand I'm biased towards GDC.  But I have to admit that on a second tasting, it was still just as inspiring and pioneering.

While strolling from talk to talk, browsing the indie mega booth, I realized something special about game making, especially in the indie space.  More game designers feel ownership over what they make compared to their VFX counterparts.  In the VFX and computer animated industry, an army of TDs and animators may have individually worked on a handful of shots in a big blockbuster film, but the film is owned by only a few at the top.   It's much harder if not almost impossible to make a living on small team film projects.  But the indie gamer market is more forgiving of well designed, small team made games.

I was also surprised to note how many women were at the conference.  As much as I complained about the long men's room lines during my last year blog,  I was pleasantly surprised to see more women in the writing, designing, and technical talks I attended.  I'm a big fan of more women in technical efforts and was glad to hear my fanaticism reaffirmed with actual scientific studies at Jesse Schell's talk.

This year's highlights:

1. VR is officially an epidemic

As much as I talked about VR being the hot topic of last year's GDC, this year VR has reached the next phase of infection, being a presence at almost every major booth at the exhibit.  Every game engine company, every big game house, and every VR hardware maker had either an Oculus Crescent Bay prototype or another up and coming hardware in their demo areas.  As a content maker, I can't be more excited.  

Also, any chance to here John Carmack talk is always worth the price of admission.

2. Storytelling in Games

On Monday/Tuesday, I attended a lot of talks at the Narrative Writing for Games track.  I was encouraged to see how game writers think of the past few years as a renaissance in their art. Finally, the empathy felt by the player is becoming the most important part of game design.

I especially liked a point made by Stephen Hood, CEO of Storium, during his talk Computers are Terrible Storytellers - Lets Give the Humans a Shot.   He made reference to a director commentary of the 1998 film Pleasantville, in which the makers said they didn't want to shoot the movie like a 1950s sitcom (with 3 fixed camera setup), but film the movie like you were cinematically composing the world of a 1950s sitcoms.  Likewise, as a VR content maker, we shouldn't think of shooting our stories like through a camera, but think about how to create presence in a cinematic world.

The other talking point that kept coming up is the concept of "white space." Each speaker mentioned that a good story told leaves the right parts out and lets the audience fill in the blanks.  This is consistent with what we've found, that rushing a story in VR doesn't give the viewer time to discover the world for themselves.

3. Unity 5 Sequencer looks very promising

In the Production with Unity 5 talk,  I got to see a glimpse of the sequencer tool being developed on the Unity R&D team and used to make their GDC sizzle video "The Blacksmith."  My favorite description from the developer is that you can think of Unity's sequencer as a way to design a generic event graph that can change over time.  This holistic approach to authoring timeline based action is much more flexible than what we've experienced with UE4's matinee.  Matinee is a powerful tool, but there are times when we're reminded that it is a limited tool mostly designed to make cut scenes in Gears of War.

4. Telltale Games doesn't care if you don't call their work "games"

At the Tuesday talk When Story is the Gameplay: Multi-Genre Writing for Telltale Games, I smiled to hear Kevin Bruner say that he doesn't care if their work is called games or not.  I especially loved a moment when one of the panelists talked about how most games use cut scenes and story moments as the carrot.  In telltale games, the entire experience is the carrot and the interactivity is just another device to create empathy.  Cool.

5. Valve VR Advanced Rendering talk was smart and open

The VR Advanced Rendering talk given by Alex Vlachos was very smart, and surprisingly very open.  Alex shared a lot of insights and facts about Valve's headset and rendering engine that could have been kept as competitive secret.  Instead, Alex shared a lot of the software solutions that will help elevate all engines supporting VR.  Here are my highlights:
  • HTC Vive uses a 90 Hz, global shutter display with a 2160x1200 framebuffer and FOV of 110.
  • Alex liked to think of the VR display measurements as Shaded Visible Pixels per Second, a great way to capture resolution, refresh rate, and overdraw requirements in one number.  As a baseline, our 1080p TVs rendering @ 60 Hz push about 124 million pixels/sec.
  • With HTC Vive's specs, the SVP/second is 1512x1680x2 @ 90 Hz: 457 million pixels/sec - (later reduced to 378 million pixels/sec by using a smart stencil mask).
  • Valve's Source engine implements a "running start" which adds about 2 ms of latency to the sensor to display timing, but takes better advantage of idle time on the GPU that is waiting for the vsync event.
  • 8xMSAA with a forward renderer seems like the golden standard for anti-aliasing in VR, a fact we've realized as well after having worked with temporal AA with UE4's deferred renderer.  Forward rendering is especially appealing with features like OpenGL 4+'s glMinSampleShading which allows us to modify shading rate on a per object basis.
  • Normal maps still work well at a distance in VR, but the most notable problem is how lower mip levels of a high frequency signal can lose interesting specular properties.  They fixed this at Valve by implementing a process where the mip-level creation of normal maps and roughness maps happen at the same time, allowing for the detail lost in normal map filtering to increase the roughness value.
  • The Valve Source engine uses a stencil mask that cuts out pixels that are heavily distorted on the outside of the eye buffers, saving about 17% of the pixel shading costs.  This may cause issues with technology like timewarp.
  • The Source engine has a highly optimized distortion mesh that only exists for non-black pixels, saving about 15% of the final distortion cost.

6. Jesse Schell punched me in the gut... figuratively

One of the final talks I attended was Jesse Schell's Game Studio Management: Making it Great.  For as dry as the subject sounded,  I found myself riveted, taking more notes for his talk than for all of my other notes combined.  Everything he said came off as charismatic, thoughtful and genuine.  And some things he said made me want to punch myself for how obviously stupid we've managed the beginnings of Story STudio.  Here's a quick distillation of the things he said and what struck me.

I. Communication as a ritual

As a small studio, we're continuously trying to work on process that is routine, provides clear over communication, and doesn't consume the team's time with meetings.  A few parts seem to be working but the hardest part in a fast moving creative environment is communication.  The common approach to manage communication is to schedule more meetings, send more emails - and it's a trap we have been fighting.  But Jesse suggested that good communication is not about meetings but designing a culture where communication happens without management.  

Jesse contended that most communication comes in subtle, difficult to measure ways born out of clustering teams working on common projects (when sometimes closing a 15 foot gap between desks can make all the difference) or giving space for unmanaged creative outlets in the forms of game jams.  I especially loved his anecdote about how his office would come together in the kitchen when the only team member who knew how to use the crappy brewing pot would make a batch.  When he wanted to encourage this "coffee time", he bought a nicer Keurig that had the adverse effect: it ruined the ritual.  It also reminded me of how much I loved the design of the Steve Jobes Building on main campus Pixar.  All meeting rooms, bathrooms, and food are located at the center, forcing cross pollinating traffic.

We've had a few rituals in our office that have faded away as the team got bigger - like board game lunches.  This idea that ritual helps in so many unquantifiable ways kicked my butt to make our team's ritual more of a priority.

II. Information cataloguing

Jesse had another great insight when he started thinking about managing information flow at his game company.  He started by listing every type of information that flows in his office, especially the info about how the team feels about themselves and each other.  He then categorized these many types of communication as fact, opinion, and/or emotion.  It was thought provoking to split emotion and opinion because if you think about it, an emotion is more fact than opinion.  If a team member is pissed or frustrated, that's a fact.

I find myself falling into the trap of thinking that a team member's emotion is a by-product, a hurdle or problem to be solved.  Recognizing that emotions are as important as the facts and opinions of your business helps you address them as the priority they should be.

III. Culture fit and the lovable fool

His final brilliant and striking point was that most successful businesses prosper by selecting the lovable fool over the competent jerk.  Even if some one is struggling, a team will rally around some one who is lovable more than they're willing to work with some one who they'd rather avoid.

I've been very proud of the team we've put together so far. We've been lucky to have not only amazing talent but really likable people.  We used "culture fit" when describing candidates during our recruiting process and I now realize that this is a bloated phrase.   In the end, the most important aspect of a good team is that everyone likes each other.  So instead of asking is some one a "culture fit", we should just ask ourselves if we like them.  Therefore it's important that the entire team plays some role in making sure that if we add a new team member, people will enjoy hanging out with them.

not about management, but a taste of Jesse's style:

The modern games industry is only 25 years old.  Film is well over 100 years.  There's something about a young industry that avoids the trappings of precedent.  I have more faith in the games industry making progressive changes in gender equality and best practices.

Imagine working in a new born industry like VR.  I'm excited to see how we evolve.