Oni2:Slaves of War/Design

Design Notes
Various thoughts on the design of Oni 2 and games in general.

Show, don't tell — Story isn’t meant to just be told during cutscenes. Put the story into the world in bits and pieces: consoles, posters….

Independent world — Stuff happens without being driven by you. Even bystanders have their own lives and personalities.

Experiment with emergence — Play with different rule sets that can lead to unexpected results; let the system express itself.

Possible value-adding features (in addition to the "bare minimum" list in the Overview):

• Alternate pathing, exploration, and optional powers. Allows player to make game their own. Optional equipment and AI allies change flow of level (e.g., a character could get captured by the enemy and held hostage, and you have to rescue them or they actually die; this would not be scripted, but dynamic). Mai's powers develop in a direction based on play style (e.g., stealth-focused, power-focused, etc.).
• Realistic AI. Enemies are unpredictable and inventive, understand your intentions, and react to your situation in battle.
• Levels change with time of day. Security in level differs, use of cover is altered; requires truly dynamic lighting.
• Memorable boss fights. Every fight is an event with stages, set pieces, and distinct style.
• Full interactivity. The player is never restricted by non-interactive cutscenes, and expresses her personality through her play style. See "Free agency" below for details.
• Procedural outlying world. This would allow us to avoid worrying about invisible walls or real walls which only serve to pen in the player.
  

It might be interesting to build on the notion of Mai’s intuition by allowing her to predict opponents’ behaviors or combat moves. It could be that her intuition is only at 100% once she has transformed, because her Imago form will have much keener senses. Until then, she can only form conclusions based on the data that her limited senses are taking in. Perhaps a character who knows how the Daodan works points this out to her during a fight where she is starting to predict his attacks.

If we’re going to emphasize the importance of trust or cooperation in the story, and even if we’re not, we should probably open the game with Mai in ruins or wilderness. The levels are empty at first and gradually she encounters people. We can use bad weather to make things feel lonely and oppressive at the beginning.

How will we deal with the cognition limit?

Can we use infrasound to affect the player’s mood?

Design levels with the current amount of tension in the story and mission in mind. Tension can alter the openness of a space, the music and sounds, the number of civilians present, and even enemy types. Which is not to say that open spaces are automatically less tense; a clearing in the forest is pleasant in the daytime, but a campfire in a clearing at night is tense because you can’t see what’s in the darkness around you.

Set to music like BG&E’s “Return Double HH’!”, picture a stealth level from a top-down view with a camera that auto-zooms and auto-centers to show you the nearest enemies, and shows their vision cones like flashlight beams. Go here for info on human vision.

AIs are people too. Let them be distracted by things near them while walking casually (Oni already has AIs look in random directions when not alert). Make them react to sounds after a short delay while they think, “Did I just hear something?” and slowly stop walking, then turning to look thoughtfully at the source of the sound, or around in general. Make sure to restore the idle-to-alert loading-weapon animations.

Crowd actions can be scary when they’re hostile or unexplainable. They will also be unexpected in a game that rarely has many people on-screen at once. Imagine something like a zombie parade or The Big O’s March of the Foreigners while they sing their national anthem.

There should be times when the player is impressed by scale. What emotions can we evoke with scale? Awe, at the size of a level. Loneliness, at the emptiness of a space. Fear, at the number of enemies. Imagine a line of enemies, giving off a light of some kind, winding their way down a hill towards the player. Imagine hordes of enemies rushing over a hilltop at you (we don’t have to be able to draw hundreds of enemies; we can use simple box models as stand-ins for distant ones).

Allow exploration — Don't want world to feel linear, even if it is. See this famous pic about FPS level design -- we want to avoid that feeling.

Create a feeling of anticipation in players to get to the next level. This can be done through making a plot that they want to advance, as well as by providing thumbnails for future levels. These thumbnails would be large, vivid and carefully composed. They could be a fading slideshow of different parts of the level, or even a short, seamlessly looping movie with ambient sounds from the level.

Allow veteran players to do more or find more stuff in a level. Allow players who train more to unlock moves earlier. See this article for reference points.

The training level can start with the ABCs from Oni, perhaps even the same level, and if the player plays it, she hears Mai reminiscing about Shinatama. But she can also skip straight to later “lessons”; these lessons can be taken in-between any mission. If there is a voice guiding Mai and instructing the player how to throw, it would be a great moment if they tell the player that you cannot throw someone who is facing you unless you knock them off-balance first, and then the player is instructed to try it, just to see that it’s not possible. Mai then successfully throws the person, surprising the voice in your ear. She’s strong enough to do this, but the average character class used in MP cannot (see “throw spamming” notes below). Shapeshifter should be allowed in VR training. The premise can be that Mai can see through the eyes of her opponents to better understand how they fight. A message can be displayed at the bottom of the screen saying “Shapeshifter enabled. Press F6 to cycle back, F8 to cycle forward, and F7 to return to Mai.” Later on, we only have to say “Shapeshifter enabled" to remind the player that they can do this.

Difficulty should oscillate — An enemy gets easier as the player's ability or Mai's power progresses, then you introduce harder enemies, then the player gradually gets better and those enemies become easier.... This allows the player to sometimes feel powerful and sometimes feel challenged.

Don’t forget that Oni made a point of putting Konoko in different outfits. It would be a glaring oversight not to do the same in Oni 2. Perhaps the costume changes can be more justifiable this time around. Maybe armor actually armors her this time around?

Music can add a tremendous amount of mood to the setting, and speak some of the emotion for the characters. Introduce unique scores when they are most effective, such as boss fights.

Expressive characters — VG characters that are supposed to look real simply don’t — the technology isn’t there yet — but realistic characters always look their worst when they’re in cutscenes, because (1) there are close-ups, and (2) they are supposed to be “acting”, and it brings out the uncanny valley. Let’s find a way to make characters expressive in a way that feels real even if it’s not realistic.

Motion should be our emphasis. Comic books are great for depicting framed poses and also detailed art. Live-action is great for depicting actors’ nuanced expressions. Third-person video games can’t do those things because they have to control fluidly (so poses are a hindrance) and they’re mostly shot from far away (no close-ups to show off detail or expressions). So rather than focus on rich textures and detailed characters, why not go for stylized, simple designs and focus on really smooth and interesting animations instead? One example would be to have differing styles of walking and running depending on whether the player is in attack mode, is alert, is relaxed, etc.

One of the biggest problems with video games is interaction with soft bodies. Hugging is practically impossible to depict decently; even if you have two mo-capped actors, the animations will require fine-tuning, and you’ll still see either a little clipping or a case of “hover hands”. Showing someone putting on a coat seemingly requires a multi-million dollar budget. The result is that the default state of a VG char. is to stand still and move his head around and hope the VA’s emotion makes the char. feel lifelike. And yet those kinds of interactive and lifelike motions can add so much life to a char. if we can figure out how to do them. Maybe animated cutscenes are the most practical way, but if they happen in realtime (like Konoko’s yawn), they can be so much more effective.

Can we figure out a way to smoothly change a character’s speed without the use of extra keys or double-tapping? Player should be able to walk, run and dash freely according to how she is feeling about the space she is in. The analog thumbstick, when accurate, is a good example of superior speed control. The scroll-wheel can probably substitute for that on PC (see Intel Mac build, where this is an optional feature).

When movement is mapped to ESDF or RDFG instead of WASD, it gives us more keys to the left (in addition to those on the right) for the player to use. It may feel foreign to a WASD gamer at first, but the tactile bump usually found on the ‘F’ key can be used as an anchor. Of course, controls will be customizable, unlike a certain other game….

Remember to vary the tone, pace, and mood of a level.

One way to emphasize the usefulness of a group is to have them move together in coordinated ways, and perhaps create combination attacks with Mai that make the player feel like she is glad to be part of a group.

Your allies can show personality, especially the named characters, by the AI showing initiative — examples would be attacking the enemy proactively, recommending a plan, laying their hand on Mai’s shoulder, talking to you in the midst of battle, and protecting you when you’re hurt.

Getting to know your allies so you can choose them for missions requires having an off-mission space where you can freely interact with the NPCs, like in Deus Ex, but better. This picture is a great reference for changing to a fixed POV that allows you to better see the NPCs while off-mission. It might be a good idea to give a player a reason to choose an ally that has to do with that character’s interests (they want to go on that mission for personal reasons, they need to get away from the base for a while to do a secret errand, etc.). That opens up the player’s ability to grow attached to the characters.

Taking along allies should make the mission easier, but could also come with a warning that the enemy is more likely to detect a group and respond accordingly, whereas going solo will allow stealth play.

Mai needs to feel powerful in combat — more so than in the first game. Daodan overpower mode in particular needs to have a palpable degree of power to it. We may want to ramp up Mai's strength over time, throwing enemies further back with final blows.

The player should feel connected to Mai — it’s problematic that the camera is always over Mai’s shoulder; you never see her face except in cutscenes. We should find a way to counteract this — maybe move the camera during special moves so we can see her face, or move the camera from over her shoulder to in front of her during cutscenes, then move it back (as opposed to cutting). It would be even better if somehow we can show her face during gameplay, but that seems impossible.

Systematize the abilities of characters who use the Chrysalis, etc. like shounen series systematize their "jutsu", "kido", "Devil Fruits", etc. Take time to reveal and extrapolate on these abilities in order to intrigue the fans, but also remember to plan ahead enough to insert certain techniques into the early game which will later be explained (e.g. look at how Naruto’s summoning technique, Shadow Clones, and Rasengan are later explained or built upon). We can also reveal that abilities that were previously ultimate-seeming actually lead to, or set up for, greater abilities (like Rurouni Kenshin’s Kuzu Ryu Sen).

Leave some mystery — Resist the urge to spell everything out in the game.

Make callbacks to Oni, even if obscure ones — sneezing, for instance (perhaps during a scene with a cyber-connection being opened, to reference the Flatline bug).

Occasional moments of quiet, especially when entering a new area, can be welcome. The point where you are going to enter a building, for instance, with the night wind blowing leaves through the parking lot; or a tower wreathed in fog with the sun shining from behind it, casting a diffuse shadow through the air around you. Those moments actually add life to the world because they feel like more than just invisible backgrounds for fight scenes.

Variety in combat is essential, both within fights and also throughout the game. Unlocking new moves over the course of the game doesn’t always make sense when your character is experienced from the start. We could introduce new Daodan abilities over the game, but that makes game replays annoying for the player who was used to having all the end-of-game abilities. We can either take the simple approach and offer a New Game Plus, or else we can give the player most of their abilities up front and try to create changing environments or sets of enemies over time that require different kinds of skills to be put into play. All this being said, new content over time is probably less important than just making sure that the basic mechanics are fun enough that the player still enjoys them after dozens of hours.

We should have some new moves, of course; not just different punches, but new specials. If Oni’s signature move was the Lariat, what would Oni 2’s be? Maybe something like this (that’s a crucifix head scissor-to-backbreaker combo, in case the link dies).

Special attacks can end with a cool flourish, unlike regular animations. But if the character misses with such an attack, or there are other enemies around, they shouldn’t do the flourish.

Alternately we could implement a stamina system where bigger moves are more tiring. This could also be used to prevent players from using annoying “runaround” tactics in MP. Throw spamming can be prevented by requiring an opponent to be worn down before throwing them (the equivalent in MMA of getting inside someone’s striking range to grapple with them), which could use the same stamina gauge.

Alternately, throw spamming can be prevented by simply requiring one or two stunning moves to be landed first. This is only realistic, as a person cannot be thrown when they’re standing still and are balanced. Back throws can still be allowed without a preliminary move.

Some characters knocked onto the ground may be able to defend against a ground attack by raising their arms to defend their torso. Others may be nimble enough to roll out of the way. Bigger characters just get the wind knocked out of them and remain vulnerable.

Remember to preserve Oni's fun interactions with civilians and the entertaining VOs they have.

The use of color should express emotions or some kind of ascending ranks. We should look to Japan for our color sense.

If we think of people as animals, how do we direct the players and manipulate them with our environments? Cf. fear of the dark, etc.

What makes for a catchy idea? What made Portal players rave about the Companion Cube, the cake that was a lie, and “Still Alive”? I would argue that all of those were so memorable because of their whimsicality, their care-free incongruity.

Some music is much more powerful when listened to in the right season/weather, so perhaps the same is true for visuals. Can environments and weather in-game be designed for the season in which that level is supposed to be? If we are implementing the feature where levels change by time of day, we can take advantage of this too with track variants (different instruments, tempo, etc.).

Game designers should ask themselves, What is my game teaching people? All games teach rules to people and therefore have implicit messages, a "meaning of life" if you will. This is not necessarily connected to the actual story, and may arise from the mechanics instead. See this article for details.

It should be possible for the player to grow closer to each ally (member of your AI team) through actions which will endear them to you. If there is a betrayal later, it won’t be the one you are closest to, because they’ll be too fond/respectful/admiring of Konoko to betray her. Each member would have their own reason to support the group and a potential reason to betray the group, so a member who you neglect is more likely to be the traitor. However winning over one member might hurt your relationship with another who dislikes that person, so you can't simply max your relationship points across the board by grinding through trust-building actions with each character. Each character has specific traits that can benefit the group more if you bond with them. One might be timid, but with a sharp sense of humor; he might be cowardly in battle but a good hacker. Another might be hard to like, and alienating to the other members, but a strong ally in battle. Given enough variety in the team, different kinds of players will have their own favorites. Look to Persona 3 or later for an example of a deep relationship system, and look to XCOM 2 for an example of getting players to invest in the soldiers they command. Also see 2002 game The Thing, where a low level of trust from a teammate restricts the actions they will perform, and you have to earn their trust back.

If we have the resources to spare, it would be fun to add something really exciting to a level but direct little attention to it, and wait for players to find it. Say they exit a base, and notice a stadium nearby with sounds of distant cheering, and upon entering it they find out they can participate in a fighting tournament. (Imagine that music like “Yo Pumpkin Head” from Cowboy Bebop is playing, to set the tone.) This kind of surprise will create a joy of discovery and lead to the sharing of these discoveries with the community.

Should we use a standard respawn mechanism where the player starts again from the last savepoint after dying, or something more interesting? Examples of games with alternate respawn systems include:

• The spirit-walking mechanic used in Prey and Legacy of Kain: Soul Reaver.
• Perma-death as seen in Escape Velocity or certain RPGs.
• Respawning with continuity as in roguelikes, Planescape: Torment, and Shadow of Mordor, where your character is already dead from the beginning, and “dying” simply costs him three days before he rises again, during which time the enemies regroup and adapt.
• Rewinding, as in Prince of Persia: Sands of Time.
• The buddy assist in Far Cry 2, which saves you after your first death but causes your buddy to enter the firefight, which means he can be perma-killed from that point on.
  

One of the problems I have with traditional save-anywhere PC games is that they take the player out of the moment and reduce the penalty for dying in proportion to how frequently one saves, leading to some players (like me) spamming the quick-save button. It would be interesting if dying meant waking up from a bad dream on the morning of the mission. This would promote immersion but be controversial among gamers used to saving anywhere, as it means even eliminating the fixed savepoints of Oni. It would probably be a good idea to limit this feature to the Hard difficulty setting.

Should cutscenes be 2D animation? This will avoid boring the player and allow for more artistic expression while also alleviating technical issues with in-game cinematics. Additional info and memories that the player may be interested in (in-game flashback cutscenes) can be found under Diary. The main argument against this feature is cost, as good hand-drawn animation is expensive.

Dream sequences can presage the development of Mai’s powers with the use of symbols, like animal eyes hinting at a new night-vision power.

The console in Oni that talks about operating on a Daodan symbiote says that the skin repairs too quickly for them to do anything. You’d think, therefore, that in Oni Konoko's health would have regenerated constantly, and that this would happen even more so in Oni 2 if she is closer to her Imago state. If we incorporate regenerating health into Oni 2, we should make sure there’s a cost. Maybe enemies are alerted by your gathering of Daodan energy to heal. If you have AIs assisting you, they can offer to surround you protectively while you heal. Healing should require holding down a button that prevents you from moving for a while, and dims your awareness (most of the screen around Mai turns black).

This also leads to a natural question about hypos. They were never realistic in the first place, or they wouldn't heal people anywhere near that quickly. Presumably they would not be needed in Oni 2 if Mai can auto-heal, but we might want to be able to use them on AI allies, which poses the dilemma of having two competing methods to heal Mai, with hypos being far more convenient. A nice solution to this is found in an old suggestion by a forum member, which is that hypos only have that healing effect on Mai because their sedatives suppress her consciousness, which in turn normally suppresses her Chrysalis. This would also explain why Mai heals under the surgeon’s knife but not when she’s injured in-game: because she was obviously sedated while they operated on her! And as far as the subject of hypos restoring the same amount of health for NPCs, nothing in Oni explicitly says that hypos dramatically heal regular people like they do with Mai, and AIs cannot use hypos, so we never see this happen in-game unless we force it with BSL. So if we ever remade Oni, we would only have to slightly rewrite some text to make it clear that normal people use hypos to recover from injuries over a much longer period of time, and that symbiotes react much differently to them. Then, in Oni 2, Mai can be more in touch with her Daodan, or learn to be at a certain point in the game, after which she can auto-heal. At the same time, hypos may become too dangerous to use by that stage in her progression if she is trying to avoid Imago.

The term “animé style” means different things to different people. Below are some defining attributes that we could try to include in the game:

• Clear hierarchies; the main good guy is apart from the rest, the bad guys are all numbered (and often introduced by distinctive silhouette).
• Dramatic posing (see "motion" and "special attacks" points above).
• Large differences in strength; rarely does a fight feel like it’s equally balanced; even if it is, the opponents are leapfrogging each other in power level. Sudden revelations of hidden powers/abilities.
• Weighty, high-impact physics; things that make you go “whoa”.
• Fairly generous use of slow-motion and close-ups for reaction shots.
• Observers to a fight serving as commentators.
• “Specialist battles” between characters high in strength, strategy, or speed.

Free agency

Cutscenes are an inevitable part of a game, due to the need for dialogue and exposition. In the industry, they are often referred to as NICs (non-interactive cutscenes). Some games allow you to choose conversational options from a tree, which makes the cutscene semi-interactive. Half-Life sets a low bar for "interactive" cutscenes, allowing you to move freely about a small area while someone talks to you, however the conversant never seems to notice anything you do during the interaction, even if you bounce a paperweight off their face, and weapons are disabled during these NICs. Even Deus Ex, which gives the player lots of flexibility in how they carry out a mission, uses cutscenes to force the player-character to behave a certain way (though some cutscenes have dialogue trees).

The idea behind "free agency" is to never take away control from the player, allowing her to alter the way a conversation or other interaction plays out with her actions, not just by selecting pre-made sentences from a list. If an NPC wants to talk to Mai, the player can ignore them, listen to them (with the push/pull response options described below), or she can even punch the conversant in the face or draw her weapon and use it. An aggressive choice might lead to a fight, or change the discussion in another way.

Rather than the player mentally checking out during cutscenes, or looking for a way to skip them, the interactions with NPCs will become part of the fabric of the game. In some cases, this will create an exhilarating sensation akin to a rollercoaster, as they realize that anything they do (or don't do) will affect the outcome of the scene* in real-time. NPCs feel more real because nothing is strictly scripted and the same scene can play out differently based on player actions.

(*Since the cutscenes will no longer "cut" (move the camera, take away control), the most logical term for any part of the game, whether there are interactions with others or not, is simply a "scene".)

The key in implementing this feature is to think of everything the player can do, and prepare accordingly. Lots of alternate dialogue is needed for each conversation, but within a clearly defined scope, because there are only so many actions the player can perform. Scenes will only have a few possible outcomes, and the story will be written with the need in mind to funnel the player into certain possible actions. This will probably feel quite natural to the player when the scene is written properly.

Since verbal freedom is needed as well as physical freedom when interacting with an NPC, the player will have two buttons for spoken replies, besides the general controls of the game being available to her. These buttons will not equate to “yes/no”, nor to "good/evil”, but to "push/pull". These buttons will encourage the player to think about the effect her response will have on her relationship with each character, and also to consider how real-life conversations work; we often respond to someone the way that we do out of a desire to push them away or pull them closer.

If we are going to feature a lot of AI allies in the game, then we also need team command controls. These will be just as important for the free-agency concept as for basic tactical purposes. For instance, one of the biggest uses of the free-agency concept would be deciding not to attack an enemy, or deciding that someone is the enemy. If you're attempting to spare someone, but your team is still attacking them, it spoils your choice. So you need a button to tell them to stop attacking a target (though if they don’t like you, they may not listen — but that depends on the relationship you've built with that character, and who you chose to bring along on that mission).

One issue to address from a storytelling angle is how to show events that don't involve the player. If we're cutting to a completely different location to watch other characters interact, how can we hold to our commitment to full interactivity? One way to bypass the expectation that the player will have control of any scene is to actually depict those remote events as animated cutscenes — by which I mean traditional hand-drawn animation. The change in the medium of these actual cutscenes will make it abundantly clear that the only control the player can expect to have is the ability to skip the cutscene. A second approach is to commit to the idea that the player only experiences the world through Mai's eyes; anything that happens elsewhere is either unknown or has to be relayed to the player. The downside is that this leads to "telling, not showing" when it comes to events that would be much more interesting for the player to see for herself.

One of the big selling points with full interactivity is that allowing more player freedom leads to greater replay value — even more so when it’s not a simple matter of playing as “good” or “evil”, because it will encourage the curious player to try different things on the spur of the moment without feeling like she knows what the results will be. Although free agency requires creating extra assets for branching scenarios, which will reduce the length of the game (assuming the budget remains the same), there will be great potential for surprising the player with how the game reacts to her, which will cause a lot of discussion and YouTube videos about the interesting experiences people are finding (like my surprise in discovering, after years of playing Oni, that you can interrupt an NPC while they're talking to you and they'll take offense — except on a much larger scale than that). At that point, overall game length decreases in importance, and in fact could even be an asset if its shorter length makes people more inclined to do additional playthroughs.

Relevant links:

Video Games are Dead: A Chat with Storytronics Guru Chris Crawford

Provocative interview with Chris Crawford on interactive storytelling.

Alive Tiny World: Deus Ex: Life, Liberty and the Pursuit of Happiness - or as close as we'll ever get to it

On what made Deus Ex almost unique in the world of games.

Gamasutra: Tears in Rain: Remembering the Blade Runner Game

On the subject of the player’s actions allowing the story to play out differently each time, the Blade Runner video game is famous for this, even swapping roles for NPCs based on the player’s choices.

Emily Short: Tightening the World/Plot Interface, or Why I Am Obsessed with Conversation Models

Emily talks about exactly the kind of challenges I have been contemplating for the Totally Interactive concept. I think that Oni 2 would naturally avoid some of the challenges that she faces with a text-based interface.

“	The real question: is game design a hard science (study of things, forces, etc) a social science (study of people), or a humanity/liberal art (analytical/critical study of people)? I'd argue it encompasses all three.	„
—Comment by Lars Doucet