The Making Of Gunstar Heroes, Treasure's Mega Drive Masterpiece

Gunstar Heroes, Treasure’s first release on the Mega Drive, is a wild, chaotic two-player run-and-gun game that stands out as one of the best action games of the era. It is also, according to Masato Maegawa, Treasure’s representative masterpiece and the game that put the company on the map.

Gunstar Heroes was created at the café meetings the fledgling Treasure team had been having around Tokyo, from before the time Maegawa first visited Sega. The team had decided on the genre of action shooting, and the design began to take shape during these café meetings. According to Maegawa, the game, tentatively titled Lunatic Gunstar, was centered on the concepts of “exhilaration” and “anything goes.” Without access to an office or development equipment, the team focused on fine-tuning the design proposal for Lunatic Gunstar.

Maegawa visited Sega’s consumer software division with the design proposal in hand during the second half of 1992. He met with a producer and showed him the proposal, but the response was not what Maegawa had hoped for. The producer, looking through the proposal, seemed confused at what the game was trying to be. He ultimately rejected the idea for Lunatic Gunstar, saying it would never sell. The situation likely gave Maegawa flashbacks to his time at Konami. “It was much easier at the time to get an original game proposal approved than it is now,” Maegawa said in 2004, “but since we were a new company with no previous record, we couldn’t get approval.”

At the meeting, the producer suggested that Treasure instead develop a game based on the McDonald’s license that Sega had just acquired in the U.S. Maegawa agreed to the arrangement, although he was disappointed that Treasure’s first game would be a licensed character-based game. Nevertheless, the decision was necessary to establish the company’s reputation and bring in its first revenue.

Why did the Sega producer reject the Lunatic Gunstar proposal? Maegawa explained: “The thing with proposals for action games is that you can never tell if the game will turn out good. Nobody looked at the proposal and just knew it was going to be good or sell well. You have to actually play the game to know if it’s going to be any good. And that’s especially true for action and shooting games.” Although the Lunatic Gunstar design proposal closely matched the final release, it arguably did not convey the technical brilliance found in the finished game. The only way to convincingly demonstrate the game’s merits was to develop a playable prototype.

Maegawa made the decision to continue with Lunatic Gunstar. He was certain that once the Sega producer saw it in action, it would be approved. Two teams were formed within Treasure: one to begin work on a McDonald’s game, and the other to develop a working prototype of Lunatic Gunstar.

Building The Prototype

Once he had signed the contract for what would become McDonald’s Treasure Land Adventure, Maegawa was able to use the advance payment to rent office space and purchase development equipment. He found a seventh-floor office in Ueno, Tokyo, just east of Ueno Station and the famous Ueno Zoo, on a street known for its numerous motorcycle shops.

Treasure followed Konami’s style of development team structure, with two programmers and two graphic designers forming the backbone of each team. Rather than having someone in the role of director, the team generally made creative decisions together under the guidance of a team leader. Sound effects and music were provided by members of the sound development team, and development support was available from others in the company if needed.

The members assigned to Lunatic Gunstar were programmers Mitsuru Yaida and Hideyuki Suganami and designers Tetsuhiko Kikuchi and Hiroshi Iuchi. At Konami, the two programmers had worked together on Contra III: The Alien Wars, and the two designers had worked together on the arcade version of Bucky O’Hare. Their experience with 16-bit hardware made them a logical choice for what would hopefully be Treasure’s flagship title.

Programmers Yaida and Suganami were exposed to many of the design principles that would shape their work on the Lunatic Gunstar prototype while they were developing Contra III at Konami. In some regards, Lunatic Gunstar was approached almost as a continuation or evolution of Contra III. Speaking to the publication Continue in 2001, Maegawa went so far as to say, “Many people see Gunstar Heroes as our defining work, and since the programmers on Contra III and Gunstar Heroes were the same and both games share a lot of similarities, Contra III can be considered one of the origin points of Treasure.”

Character designer Kikuchi, who was participating in the planning process for a game for the first time, was aware of the influence of Contra III on the Lunatic Gunstar proposal and sought a way to make the characters distinct. “Initially, the round-faced designs featured in the game were not how I typically drew characters,” he told Japanese magazine Beep! Mega Drive in October 1994. “However, the game was in the same genre as a certain Contra game, so I thought it would be good to challenge myself and take the character design in a new direction.” His characters had a distinct Japanese feel to them. The designs were lighter and more playful compared with the grittier, Western-inspired designs of Contra III.

One of the challenges background designer Iuchi faced in designing this world was working with the Mega Drive’s limited colour capabilities. The Mega Drive’s four colour palettes were divided up so that the game’s background plane and foreground plane each had one dedicated palette of 15 colours. Iuchi had to create rich, colourful landscapes using this limited number of colours, preferably in a way that did not emphasize the hardware’s limitations.

Speaking to Famitsu in November 2018, Maegawa commented:

If you want to talk about challenges with the Mega Drive hardware, the low number of colors and the low amount of RAM are two of the main ones. However, they can be overcome through techniques such as carefully dividing the colors among the backgrounds and sprites, and programming the upper and lower halves of the zako enemy characters separately. Coming up with ways around these limitations was one of the joys of development.

Yaida, meanwhile, was handling the prototype’s main system programming. This was the programming of the behind-the-scenes aspects of a game, such as the game engine, sound drivers, controller input, graphics decompression and loading, memory management, and so on. Speaking in 2006, Yaida generalized the role in the following way:

The system programmer, to phrase it elegantly, is the unsung hero of game development. They’re not involved in the more exciting, visible parts of development, but they’re responsible for the whole program code. The system programmer is the unfortunate one who is called in the middle of the night when a bug is discovered in the final code.

The programming of the player characters was also handled by Yaida. “It’s important that a game’s actions fit the player character,” he said. “The actions should never be standard or ordinary.” The Lunatic Gunstar team had already brainstormed an extensive list of player character actions. In addition to the standard shooting, running, and jumping actions, the design proposal included jumping attacks, wall jumping, sliding attacks, punching, throwing, crouching, and hanging actions. Yaida added all of these to the prototype, but he did not stop there.

Both Yaida and Maegawa emphasized the importance of good player actions for creating a successful game. “The player actions should always be finalized at the very end of development,” Maegawa said. “If you firmly decide them at the beginning, the game will never evolve into something better.” By having an “anything goes” style of development where all of the team members were free to make suggestions, the number of player actions gradually increased as development progressed. Maegawa continued:

When you’re in the middle of development, you’ll have the sudden thought: “Shouldn’t we insert this action, too?” So, more and more actions were added: jump attacks, dashes, and those kinds of martial arts moves, and then mid-air throws, big jumps, and even blocking. There’s a lot of “How about this?” “Yeah, that’s good!”

Kikuchi, meanwhile, was working on a similarly involved aspect of the game: the actions of the enemy zako characters. Zako means “small fish” in Japanese and is used in video games and other media to refer to weak, unimportant characters, such as the small-fry enemies that fill most action games. Kikuchi deliberately named the enemies in the Lunatic Gunstar proposal “zako,” reflecting their role in the game. Unlike many games where it was easy to overlook such enemies, however, Kikuchi wanted them to stand out. In 1996's Rakugakicho Compilation, he developed their back story:

Zako. They’re just as their name suggests. They’re androids (it’s not good to be killing people, after all) that have had just their brains cloned. The captain that sometimes appears is a copy of an excellent soldier, while the zako are copies of this copy, so they suffer from a bit of quality degradation.

Kikuchi next worked on the zako design. He described some of the challenges behind the design, writing:

It’s very difficult to design zako. They should be simple yet clearly show personality, they should be versatile, and they should be mass-producible. The designs of the Zaku and Leo [from Gundam] are truly wonderful in my opinion, as are the enemies in the Super Sentai series. I have to study them.

He then developed the actions of the zako characters. This is where they truly shined. Kikuchi created and animated dozens of actions for the zako, many of them reflecting their rather silly nature. “The zako have more actions than the player characters,” Kikuchi wrote, “although the player characters have more animation frames and shooting actions. If you leave the zako alone, they make different poses, and if you keep leaving them alone, they run away!” The zako walk, run, jump, kick, slide, crouch, throw bombs, and grab the player, among other actions. Incredibly, the zako actions filled around 400 graphic tiles in the Mega Drive’s VRAM—nearly 20% of the system’s 64 KB.

Maegawa praised the final design of the zako:

The characters that [Kikuchi] designed all have unique movement patterns. They are really defined by how they move, and this movement is quite detailed. For the zako characters, even though they are zako, they have quite a lot of movement patterns. If you watch the screen carefully, you’ll spot them doing some really silly things (laughs). You might think, since they’re zako, all they have to do is walk, shoot, and get blown up, but that’s boring. It’s something you can’t see from still images of the screen, but the zako really do move a lot. I think this is one of the points that players really appreciate.

The prototype was taking shape. The team had transformed the player characters, the zako, and the stages as described in the proposal to a playable format. One major element remained: the bosses. Hideyuki Suganami had been using the downtime after leaving Konami to develop many of the concepts that would drive boss design in Lunatic Gunstar. These concepts showed a refinement of the work that he had done in Contra III, and more than ever, they showed the fusion of his artistic vision with his newfound technical skill.

Suganami was drawn to the possibilities that existed within the use of multi-jointed characters. These were characters formed from small, repeating linked segments, most often found in shooting games in the form of a serpentine boss that flowed smoothly around the screen. Suganami was interested in the question of how repeating sprites could be linked together in more complex ways. There was a flavour of this exploration in some of the Contra III bosses that had multi-jointed limbs, such as the leg of the giant tortoise Taka, but the movement patterns were still simple. An advantage to using multi-jointed bosses was that they could be animated in complex ways through the program code alone, and this animation could be very smooth—at the same 60 fps at which the game ran. This would obviate the need to create memory-heavy sprite-animated bosses that moved less fluidly.

As he began work on the Gunstar Heroes prototype, Suganami became enamoured with the Mega Drive hardware, especially its Motorola 68000 CPU. “The 68000 is great,” Suganami told the readers of Beep! Mega Drive in 1993. “It’s much better than the 65816 [the SNES CPU]. Its instructions are easy-to-understand, and it’s fast. When you do multiplication, you don’t have to jam a bunch of NOP instructions in there. The Mega Drive is truly a wonderful machine with a CPU like this.” On the SNES, it was necessary to include a series of four NOP (no operation) instructions—which do nothing but run two CPU cycles each—when using the unsigned multiplication function of the hardware.

Programming the movement of a multi-jointed character was tricky, since the positioning of each segment relative to its adjacent segments had to be known each frame. There were two general ways to do this, either through massive tables of data (known as look-up tables) stored in memory or through real-time calculations. Look-up tables were the simpler way of doing things. The relative positioning of each segment could be computed in advance, stored in the ROM, and then simply referenced when the character was on screen. However, Suganami was firmly committed to doing all of the movement with real-time calculations. “If you store all of the movement patterns as data,” he said, “then yeah, the processing is fast and the appearance is probably quite realistic, but that data is going to take up a huge amount of memory. I am firmly against this above all else. Multi-jointed characters have flavor because their movement is calculated in real-time!”

Although both the SNES and the Mega Drive CPUs performed multiplication at similar speeds, the Mega Drive was capable of multiplying two 16-bit values together, while the SNES was only capable of multiplying two 8-bit values together (or, using its graphics processor, an 8-bit value and a 16-bit value). This may have limited programmers who were looking to do complex movement calculations using multiplication on the SNES.

Suganami focused his efforts on the boss of the second stage, the Underground Mine. On this stage, the player rode a fast-moving minecart that could switch between riding on the floor and ceiling or, in vertically scrolling sections, on either of the walls. Jumping between the minecart tracks was an interesting mechanic that added variety to the gameplay by increasing the player’s range of movement, and it also set up one of the most jaw-dropping boss fights of the generation: Seven Force.

The concept of Seven Force came to Suganami shortly after quitting Konami, when he suddenly found himself with a lot of time on his hands. Ostensibly to research ideas for the new game he would be working on at Treasure, he had been watching a lot of television anime. “There was this one show called The Brave Fighter of Legend Da-Garn,” he said in Beep! Mega Drive, “and it had a robot that changed between seven different forms. I thought, this is awesome! And I decided to make my own seven-changer, but as a multi-jointed character.”

The initial Lunatic Gunstar design proposal featured a slightly more modest boss called Five Changer, which was piloted by the main characters’ former friend and ally, Green. As Suganami worked on the prototype, he revised the boss to one tentatively called Seven Changer. In updated design documents, Seven Changer was described as the game’s ultimate selling point—a boss that could transform between seven forms. Eventually, Suganami would settle on the more original name of Seven Force.

Suganami spent an entire exhausting month on the process of designing and programming Seven Force. He recounted the ordeal in Beep! Mega Drive:

After I decided to make Seven Force, then the painful days really began. I spent a long, confused month on it and it was finally finished. A whole month! Was I crazy?! I spent an entire month of our precious development period—as small as a sparrow’s tears—developing a single boss! Well, whatever. You’ll never see anything like it in another game. The way it moves is quite a sight. I wonder if I’m the only programmer in the whole world who has expressed the movement of a tiger as a multi-jointed character like this.

The most remarkable form of Seven Force was the humanoid Soldier Force. Soldier Force’s movements were striking for their realism, and the effort that Suganami put into this particular form clearly showed itself. He described discovering the movement pattern as follows:

I really got stuck on the running motion of the humanoid figure. I tried so many things but they never looked like a person running. Then, one day, I had to run home through the rain because I forgot my umbrella. In the middle of that heavy downpour, it suddenly hit me: This! This is the motion! Of course. From the beginning, I should have just tried running myself. But that moment was incredible. I owe my thanks to the rain. I was running through the downpour, laughing with joy. Being a programmer requires a lot of subdued, quiet determination, but the joy that you feel when the solution to a difficult problem flashes through your mind is unlike any other job.

In 2022, Maegawa recounted his memory of Suganami and Soldier Force:

"I’ve figured out how to create this incredible motion!” Those were the words Nami blurted out one day right as he entered the office. He immediately began working on the movement for Soldier Force. Apparently, Nami would sometimes get ideas while dreaming. That’s probably because, during game development, even when you’re sleeping, you’re still unconsciously thinking about the game. But I was truly impressed by the magnificence of his ideas. He had the programming sense of a motion designer, and it was eye-opening to witness. I guess that should be expected, given that he was originally a designer and not a programmer.

With the addition of Seven Force, the Lunatic Gunstar prototype had reached a point sufficient to show off to Sega. Maegawa met once more with the Sega producer. This time, the response was different. The producer was so impressed with the prototype, especially Soldier Force, that the project was approved. To Maegawa’s delight, Treasure would be able to release its first original game.

Creating A Legend

With the prototype approved, the team continued to work around the clock to bring the full game to life. The game’s title changed from Lunatic Gunstar to Blade Gunner before eventually settling on Gunstar Heroes. The weapon system—one of the game’s main selling points—was finalized. Each of the 14 possible weapons allowed for a different way of playing, which added a large degree of replayability to the game. Some weapons worked well on lower difficulties but were ineffective at higher difficulties. Some were effective against certain bosses but difficult to use against others.

The ability to select between “fixed shot” and “free shot” was also added to the options; the former held the player stationary while they fired in different directions, and the latter allowed the player to fire while running. The original design proposal only had “fixed shot” settings. The life system was also changed from the initial design proposal. Originally, the player could select between “extra lives” mode and “full health” mode. In the former mode, the player got three lives, each with 100 health, and in the latter mode, the player got one life with 300 health. This was dropped in favour of a simpler approach: one life with 100 health.

Kikuchi also designed a new boss character absent from the original proposal: Smash Daisaku. Kikuchi intended him to be comic relief—a character that “shows up occasionally and gets in the way of the main characters; he’s crazy and will throw around anything at hand—even his own soldiers” (according to the Blade Gunner design proposal). Smash Daisaku’s design was clearly inspired by Street Fighter II’s M. Bison (called Vega in the original Japanese version). He wore a red military uniform with flaring jodhpur trousers and a peaked cap in the style of uniforms worn by officers in Nazi Germany. Kikuchi even wrote “German” next to a sketch of the character, suggesting this is what he had in mind. At one point in the game, Smash Daisaku flies headfirst through the air in a way very similar to M. Bison’s special move. Kikuchi was up-front about the resemblance, writing, “This M. Bison-like move is really bad. We’re totally going to get sued by Capcom!”

Smash Daisaku’s chin featured a bushy beard sprouting out in all directions (with a bare lip), although this was often mistaken to be a coat collar on the in-game sprite. Because of this beard, the character’s original name was “Higemaru” (essentially meaning “Mr. Beard” in Japanese). Higemaru was also the name of a 1984 Capcom arcade game featuring a pirate captain with a bushy beard, and this was referenced by Capcom’s 1989 arcade game Strider in the form of a pirate boss named Higemaru, Jr. Smash Daisaku, therefore, was the fusion of two Capcom bosses—a fact that Kikuchi did not shy away from poking fun at.

The team put significant effort into the game’s first stage, the Ancient Ruins. This would be the game’s showcase, the first impression designed to amaze players and critics. In this stage, the player encounters the first mini-boss—a large balloon-launching flower called Papaya Dance—just 30 seconds after starting. This simple mini-boss was constructed from a single static background layer. Using a line-scrolling effect, Papaya Dance appears to vibrate wildly back and forth. The team nicknamed it chinchin-hana (“penis flower”) due to its amusing shape and movement.

Only 30 seconds after defeating Papaya Dance, the player ascended a pyramid and encountered the next mini-boss, Bravoo Man. It seemed like the team had added this boss for one reason: to shout out that this game was something different, something worthy of praise. It was the ultimate showcase boss—it did not fit in with the theme of the stage whatsoever, but that did not matter because it looked incredible.

Bravoo Man was a humanoid multi-jointed character composed of what appeared to be wooden crates or cardboard boxes. Its movement was fluid and smooth as it walked, jumped, and flew around the screen and made a variety of attacks (some of which bore a strong resemblance to certain attacks from Street Fighter II, of which Suganami was an avid fan). But what stood out the most was the 3D effect it demonstrated as it moved: each of the 16 boxes that made up its body and its head rotated—as if a 3D cube—in sync with the overall multi-jointed movement of the character. The 3D effect was so convincing that it was sometimes mistaken as being made of textured polygons. The reality was nothing so complicated, however. Instead, the design of Bravoo Man combined multi-jointed movement with simple sprite animation.

Bravoo Man’s early appearance on the first stage seemed designed to impress, and it would feature prominently in many previews of the game alongside Seven Force. Rarely, if ever, had a boss character with such smooth animation and wide variety of movement patterns been seen in a console game up to that point. Maegawa, though, was quick to point out that it was not that technically impressive from a programming standpoint:

I often hear words like “skilled craftsmen” and “technical masters” used to refer to Treasure, but I don’t think that’s referring to our programming skills. Actually, in the company, we would jokingly call ourselves “skill-less Treasure” (laughs). For example, with the multi-jointed bosses, we weren’t doing anything technically impressive in terms of programming. All we were doing was moving some sprites around. However, the way we displayed that movement and the ideas we put into it made it into something impressive. I think that’s what people are talking about when they say our games are amazing. If the way the characters move is fascinating, then it doesn’t really matter that our programming skills are lacking, does it?

Each of the game’s additional stages featured something unique in its design that added variety to the gameplay. Stage 2 had the auto-scrolling minecarts. Stage 3 was focused on aerial combat against jetpack-equipped zako and a flying Smash Daisaku. Stage 4 included the Dice Palace, a boardgame-like section where the player tossed a giant die to determine which room to enter. Each room in the Dice Palace featured a mini-boss, a challenge, or item pickups. Stage 5 was a zako-filled stage that required the player to blast through wave after wave of zako. In Stage 6, the player boarded a spaceship and engaged in shoot ‘em up gameplay. Finally, Stage 7 featured a boss rush that required the player to defeat new forms of all of the game’s main bosses.

Playful jokes and Easter eggs were scattered throughout the stages. The names of many of Suganami’s mini-bosses, for example, remain indecipherable, and it is not clear if they represent in-jokes or just whims of their creator. There was Curry and Rice, a mini-boss that did not appear to resemble the food it was named after yet danced around the screen in an Indian-inspired dance. There was also Melon Bread, another mini-boss named after a popular Japanese food. Again, the design of the boss itself did not resemble the food, but rather “a dog with only a face,” as Suganami put it.