GDC: Feelplus' Nakazato Details Lost Odyssey's Collaborative Process
Feelplus president Ray Nakazato was at the 2008 Game Developers Conference to discuss the collaboration between Final Fantasy producer Hironobu Sakaguchi's Mistwalker group and Feelplus to develop Lost Odyssey. It originally began as an in-house Microsoft project before Feelplus assumed a role, Nakazato explained, as he showed a trailer of the fantasy RPG.
Feelplus was established in 2005 to develop Lost Odyssey. Currently, there are 100 developers and artists on staff, many of them Microsoft and Sega veterans, and Feelplus also relied on freelancers to help develop the game. The studio is part of holding company AQ Interactive Group, a larger merger between three development studios: Artoon, Cavia and Feelplus. In addition to supporting Sakaguchi on Blue Dragon, Feelplus contributed to Yoshi's Island.
Cavia was responsible most recently for developing Biohazard (Resident Evil): Umbrella Chronicles for Wii. Altogether, the three studios have some 300 employees. Currently, their primary business is to make games for other publishers, but AQ Interactive has recently become a publisher itself, having recently acquired U.S.-based publisher XSeed.
Nakazato then explained the division of labor involved in Lost Odyssey: The project was funded and project-managed, tested and asset-localized by Microsoft. Mistwalker, with Sakaguchi and award-winning Japanese novelist Kiyoshi Shigematsu at the helm, took responsibility for the story and character design, with well-known Final Fantasy composer Nobuo Uematsu creating the music, while Feelplus devleoped the actual product.
"To develop games, or to try to get professional game developers under the auspices of Microsoft was quite difficult, and it could be that Microsoft employees are expensive. So for that reason, they decided to create an independent studio, which is why Feelplus was established," Nakazato explained. It took the team more than four years, including the earliest days of Lost Odyssey's development, to create the game.
In the middle of 2006, Feelplus developed a Japan-only playable demo. "We developed this, or finished it, around June 2006, and then gave it off to the players in December 2006. But I think this was too early, because it ended up being a version that wasn't particularly polished at that point. But we launched a playable version a year and a half before the full launch," Nakazato recalled.
The game uses Unreal Engine 3 as middleware -- and the team did UE3 integration four times, which Nakazato said was "quite a task."
For the ambitious development project, the team divided into several groups: one responsible for building the game itself, a design group responsible for database systems and AI, and a level design team. A production management group was in charge of the game's cutscenes, of which there are over eight hours, including about 40 minutes worth of pre-rendered movies with the rest as scripted real-time events.
"All of these groups would create these parts and then put them together, and then the game and the cutscenes would be put together by combining these components. And then the Microsoft project manager would be in charge of project management," explained Nakazato. Finally, a Mistwalker-Microsoft liason would coordinate among the groups.
What Went Well
"Part of this is looking back and learning my lessons as well," Nakazato said. "The game system itself is a traditional JRPG, so we had a lot of people who had a lot of experience with this and we didn't have to worry. There are so many people used to developing traditional JRPGs, that from the perspective of game design, I don't think there was a problem."
As far as storytelling is concerned, the story's amnesiac, immortal protagonist whose past is told through 15-minute short-story dream sequences are "really good," Nakazato said. "[The stories] don't necessarily has anything to do with the game itself, but thanks to them, the story and the character have much more depth."
Nakazato was also pleased with the character and creature production staff, who he says produced great-quality content on time so that there was plenty of time left to design details.
What Didn't Go Well
"I think perhaps we started too early in the project, and ended up having too many people involved at too early a stage," Nakazato said. "We could have started with a fewer number of people. Game designers had to start the design before they even had the [Xbox 360] hardware, so a lot of what they did had to start off with trying to imagine what things might be in the future. So we had to redo a lot of those things."
Secondly, the battle, adventure and cut scene components were created separately. "It's fine while they're being developed, but while they were being combined, we realized there were various issues. So from now on, a seamless development is something I should think about doing," Nakazato explained.
Nakazato also found the environment production challenging. "Part of the issue is that there were more than 300 locations that were created. And it might be that one location is very important, while another might be a room that nobody goes into. But those were built to the same quality level. Perhaps had we concentrated more on the more important scenes, we would have been able to save cost and time as well as balancing things out better, but we ended up spending a lot of energy on locations where the player would rarely go to."
Nakazato also feels the team over-used the concept art stage. "In retrospect, we didn't need to have that much details [in the concept art]. I think we overdid it a little regarding this."
Cut Scene Production
The game's visual director was a film industry veteran for whom it was his first time working in the game industry. "He imported his way of doing things to our stuff, but at the end, we had a great product," Nakazato said. He was also pleased with the hand-animated facial expression, which he felt yielded a better end result than automation. "We also decided to use English voice acting, and we created a great facial expression which is going to fit the English voice acting, in particularly eye movement."
The text-based dream sequences featuring Shigematsu's storytelling is another point with which Nakazato is pleased. "Some people are against the idea of using text on a next-gen machine, but Mr. Shigematsu's story was so great that we convinced people to follow his lead."
However, the priority list caused challenges for the cutscenes. The team broke down the game's scenes into four types: pre-rendered FMV, A events, B events and scripted events. "When we divided them we didn't have a good understanding of the quality of each scene. Maybe we didn't need to use FMV. Pre-rendered events and A events use the same models, but has a slightly different tone... but we made a mistake in that we used the real-time cutscenes, but in real time we had the two prioritizations," explained Nakazato.
He continued, "In the A event, we first created scenes and we took a video of that scene and then we did a motion capture. But in the B event, we used off-the-shelf mocaps to create the scene. But in the A event we manually created facial expressions. So we did two different methods, but we ended up that the players didn't know which one is A event and which one is B event, so they just felt that the quality of the entire movies are inconsistent. That's what we are regretting at this point."
Working With Unreal Engine 3
As Nakazato said earlier, the team started without Xbox 360 hardware or rendering technology. "Unreal 3 was at the cutting edge at the time. Also, the PC version was available before 360 existed. So we actually used the PC version before the 360 was available, and we have a great toolset."
On the other hand, "Using the Unreal Engine is also a bad thing. In Japan, those who are used to JRPG engines have a different style. They have their own philosophy. But Unreal 3 Engine also has its own philosophy. Both are inconsistent and incompatible. If we knew that earlier, we could have a consensus that we should use Unreal 3's method and philosophy... but we just applied our old philosophy." Overall, he said, the integration did not go well.
"We feel like we are porting our product to a new platform," he said. "We expected it's going to be just a little update or version-up, and if UE3 has a new version, we can use it quickly. But every time, when we start using the new version when we integrated the new version into our engine, it took six weeks; it took a lot of time and effort, and everything evolved so quickly and rapidly so integration was very difficult. Now I hear UE3 is getting more stable, but at the time it evolved so quickly it was difficult to keep up with it."
The Japanese developers, Nakazato added, had a lot of difficulty reading UE3's manuals. "It was difficult to absorb the knowledge. For example, if there were some issue or failure in the Xbox it's really difficult to understand why it happened. So the debugging and final tuning ended up being quite cumbersome as well. We do hear that it takes a very long time to load, but it actually is 3 percent quicker in loading even now compared to before."
Nakazato felt that Microsoft's milestone-based development structure was a good idea. "Those who are developing in the West are much more advanced compared to the way we do it in Japan, and Microsoft is closer to the West. There is a milestone every two or three months, and we have to deliver our milestone to Microsoft." He also praised the company's robust acceptance process and concrete and clear, agreed-upon schedule of deliverables communicated well in advance of due dates.
"After we provided the milestone, they check it very thoroughly, and they have to accept it -- and without that, our studio can't get paid," he continued. "But these acceptance criteria are very detailed as well. Because we had a three-month milestone, there was a source of tension all the time that I think helped push us along."
Microsoft assigned two program managers and a globalization program manager to development to check the game design and quality along the way. "It seems like there are way too many managers, but everyone was very cooperative," Nakazato said.
On the other hand, he continued, it was the team's first experience both in working together and working on a new platform. "Because of this, it took a very long time before we all felt a sense of unity. Because everybody is so talented, they have their own background and their own way of thinking... so there were some disagreements. Some people didn't stay in -- we started with 10 people and at the peak we had 150, so in that process, various management issues had to be resolved along the way."
Nakazato was also pleased with the localization process. "JRPGs tend not to be able to be launched globally. But we did two of them, and that tends to be not common," he said. "We continued to consult with Microsoft and decided to create a global build system."
Lost Odyssey was localized into 9 languages, and each region had a localization department to be translated and then returned for inclusion. "That didn't work very well, so we ended up having them become part of our team. The in-house tools were made to be very easy, so that the localization team could use that as well. We taught them these things and they brought them back to their own regions. And then they would use this tool and drop it into the game and check to see that it works really well in the game. They could do that locally without involving Feelplus."
The official build was done by Feelplus, Nakazato explained, but by the time it reached them the language-based assets were "quite clean. Even those areas that changed daily could be turned around quite quickly." Even the many last-minute changes could be accommodated expeditiously.
Tuning and Testing
Global bug tracking worked out well, Nakazato said, but the bugs all needed to be fixed by Japanese programmers. Microsoft Game Studios in Japan took responsibility for translating bug reports into other languages. "Thanks to that, I think we were able to work on the bugs pretty smoothly," Nakazato explained. The team used an efficient prioritization system that Nakazato said helped things progress well even if team members didn't always agree on a bug's priority level.
There were many last-minute changes during the tuning process also. "I worked a lot with Mr. Sakaguchi, but what he tends to do is give a lot of requests when things are more or less done," Nakazato said. "But we knew that even though he did it that way we would ultimately have a much better product, so we did bear with it and follow along with him."
In summary, Nakazato said, the project's weak point was the speedy growth of the team. "I think we were able to get really highly skilled people together, but it was a new team, a new platform and a new middleware. Because of this, I think it was quite a challenging project. I don't think that could have been resolved, but we've done it once -- so if we were to do it again, we could do it much more smoothly and be more productive."