Endings Are Not Success or Failure

 

Endings in most existing visual novels work like this. High affection means a happy ending, low means a bad ending. Success or failure. Pass or fail.

But real relationships do not divide into success and failure. There are relationships where both people liked each other but it did not work out, and relationships that lasted despite no initial attraction. There are relationships that fit perfectly but the timing was off, and relationships that do not fit but neither person can let go.

To express these diverse outcomes, endings must be determined not by a single value but by a combination of states.

State-Combination Endings

This is the ending structure designed with AI.

Connected — High trust + low avoidance + mutual emotion. The most demanding conditions. Simply liking each other is not enough; trust must have been built, and the player must not have avoided key moments.

Late Love — High trust + high avoidance. They know each other's hearts, but too much was postponed. The feeling of "I should have said it then." It is true they loved, but the timing has passed.

Ruin — High emotion + deep hurt. They hurt more because they cared, and the hurt is irreversible. Not because love was insufficient, but because the way they loved destroyed each other.

Passing By — Incomplete emotion. Neither side reached a decisive feeling. A good person, but the relationship simply ended as that. The quietest ending, yet surprisingly the one that lingers longest.

These four are not "good endings" and "bad endings." They are all different kinds of relationship outcomes. Depending on what combination of emotional states the player's choices produced, the outcome is naturally determined.

When I first presented this structure to AI, it asked "Aren't the boundaries between endings ambiguous?" A valid point. So we defined one core emotion for each ending. Connected is "certainty," Late Love is "regret," Ruin is "loss," and Passing By is "blankness." These core emotions become the standard for ending direction. What music plays, what dialogue closes it, where the character looks in the final scene — all of it derives from this core emotion.

Another important point is that each ending is not independent. The memory system covered in Part 15 is deeply involved in ending direction. For example, in the "Late Love" ending, scenes that the player avoided in the past return as flashbacks. Alongside the line "If only I had said it then," the moments of avoidance that the player chose are rewound. This is a structure where the did_not_ask flag and avoided_confession flag directly feed into the direction.

The Most Painful Ending

After discussion with AI, the most emotionally powerful ending was "Late Love." The reason is clear.

Ruin is painful but definitive. "It's over" is certain. Passing By is wistful but light. There was no deep feeling.

Late Love is different. They liked each other. They knew it. But they were always one step too late. And at the end, they confirm "We actually liked each other," but there is no going back.

This is the final form of "it almost happened." "If only things had been slightly different" hurts longer than complete failure. A player who reaches this ending wants to start over from the beginning. The motivation "This time I'll do it differently" is born.

The AI and I discussed the final scene of this ending in concrete detail. When I asked "How should we direct the scene where the two meet for the last time?" the AI answered "It hurts more when they already know." Both already know they liked each other. But they do not say it. Because saying it would change nothing. That silence is the core of this ending. The difference between being able to speak but choosing not to, and wanting to speak but being unable to. Late Love is the latter.

Love Is Not the Exclusive Property of Youth

After writing "the most painful ending," the phrase "It hurts because it's youth" came to mind. But think about it — youth doesn't have to be painful. There's also the phrase "beautiful youth." And getting older doesn't exempt you from pain either.

Love doesn't only happen in youth. People fall in love in their thirties, fifties, seventies. Married people waver again. Marrying for love doesn't guarantee loving until death. Love's beginning and end don't discriminate by age.

This has important implications for ending design. Just because this visual novel has a school setting doesn't mean the ending must be "graduation" or "farewell." Relationships continue after school ends. Emotions change after marriage. A "Connected" ending in your twenties might become "Late Love" in your forties. A "Passing Through" ending in your twenties might become "Reunion" in your sixties.

When I asked AI "Is there an age limit on love?", it naturally answered "No." But when AI generates scenarios, it keeps drawing school-confined romance. That's the bias in the visual novel data AI was trained on. Real love happens without school uniforms. Even if this game starts in school, the emotional design should follow universal principles unconstrained by age.

While thinking about this, one story wouldn't leave my mind.

It was from a Korean TV show, KBS's Kim Young-chul's Neighborhood Walk. In front of an elderly woman's house, several wooden walking canes were leaning against the wall. Her husband had made them. He'd been bedridden, and he knew he didn't have long. So he started felling trees. He selected sturdy wood. Sitting down, he carved each one with care. He matched the height precisely so his wife could walk comfortably. He smoothed the surface so her hands wouldn't get hurt. Then he placed them by the front door and passed away.

The grandmother said: "My husband carved these walking sticks and left them by the door for me. He must have known he wouldn't last long while lying in that sickbed. So he cut the wood, gathered it, and sat there whittling."

The show's host marveled — he'd chosen strong wood, matched the height perfectly for his wife. The grandmother said quietly: "He told me, 'Even after I'm gone, the canes will look after you.'"

To her, those canes were the most precious inheritance he left behind. "I wish you were here. I think about you so much."

I sat still for a long time after reading this.

Not a single "I love you." No confession, no event, no dramatic moment. Instead, there's a sick man felling trees, sitting down to carve, matching the height, smoothing the surface. Each time he cut the wood, he was cutting through his own grief. Each time he carved, the focus on making a good cane let him forget the sadness for a moment.

Why does this hurt more than a twenty-year-old's confession? Because decades of time are embedded in it. Every road they once walked together — now she walks alone. A man who made something to protect her in his place, then left. An emotion too large for words, contained entirely in a wooden walking cane.

The "action versus inner self" relationship discussed in Part 13 appears here in its purest form. Not a failure of expression — but something words cannot hold, poured into the hands instead. And when the line came to mind — "As the cane wears down, time wears down with it; when time's wearing is done, they will meet again" — it connected precisely to Part 18's "there is no real ending." Parting is not the end. It is a reunion without a date.

Could a visual novel create this kind of emotion? The current setting is school, but if the series continues, those characters might reach their forties, their sixties. The choices and emotions at that age will be completely different from their twenties. But the emotion system we're designing now — trust, reliance, fear of loss, the gap between action and inner self — operates regardless of age. A good emotion system should be able to hold both the flutter of youth and the farewell of a lifetime.

Endings Where Good Choices Create Tragedy

The final product of the pattern covered in the memory system emerges here.

The player made "good" choices every time. They were considerate, they waited, they respected. But the result is Late Love or Passing By. Why? Because consideration was avoidance, waiting was missing out, and respect was indifference.

What the player who experiences this feels is not simple sadness but a re-evaluation of their own judgment. "Was what I thought was good really good?" When this question extends beyond the game, it becomes the deepest experience a visual novel can offer.

This is the structure where the "paradox of caring" covered in Part 10 reaches its ultimate fruition at the ending. Seeds planted in the middle bloom at the end. Every moment the player was considerate comes back as the question "Was that really consideration?"

Transparency of Ending Conditions

An important design decision. Do we reveal the ending conditions to the player, or hide them?

My answer is "hide them." The reason is simple. If the conditions are visible, the player will try to meet them. Then choices become strategic optimization rather than emotional judgment.

The player must not know which ending they are heading toward. They simply choose according to their own judgment, and accept the result produced by the accumulation of those choices. "Ah, so this is how it turned out." That moment of acceptance is the emotion.

Of course, on a second playthrough, they can try different choices. Discovering "Ah, if I change this, it changes like that" is also part of the experience.

An interesting counterargument emerged when discussing this transparency issue in depth with AI. The AI said "If you completely hide the ending conditions, can the player interpret the meaning of their choices?" A fair point. So we created a compromise. Showing a "Relationship Retrospective" after the ending. After reaching an ending, a summary shows how the player's key choices affected the relationship. "In Chapter 4, you did not ask. That silence raised trust, but you missed an opportunity." Something like that. Post-hoc interpretation is allowed, but pre-hoc optimization is blocked.

Ending Balancing with AI

When designing ending conditions, where AI was useful was in balancing. We simulated "What ending should result from this play pattern?" across multiple scenarios.

A player who chose honestly every time -> Connected or Ruin (depending on the relationship)
A player who avoided every time -> Late Love or Passing By
A player who oscillated inconsistently -> Ruin

Organizing results by pattern like this lets you verify whether the ending conditions make logical sense. AI is very effective at this verification work. It quickly checks dozens of combinations and catches cases that produce unintended results.

There was one interesting case the AI discovered during balancing. The pattern of "a player who is honest early on but starts avoiding from the midpoint." This pattern had high trust but also high avoidance — exactly the "Late Love" condition. But emotionally, this player has the most natural pattern. They had courage at first, but grew afraid as the relationship deepened. This is a common pattern in real life too. Thanks to the AI's discovery, the Late Love ending deepened from simply "the ending for avoidant players" to "the ending for a relationship that grew then stopped." A person who stops because they are afraid of the relationship deepening. In reality and in games, that is the most common tragedy.

There Is No True Ending

Let me go one step further. In truth, all of this ending design contains a fundamental paradox.

In life, the only ending is death. Until then, you cannot conclude someone's life as success or failure. Relationships are the same. Seeing the "Connected" ending does not guarantee that relationship will last forever. Seeing the "Passing By" ending does not guarantee those two people will never meet again. The point where the game stops and the point where the relationship ends are different.

The biggest limitation I felt while designing endings with AI was this. AI creates ending conditions in order to "finish" the game. If variables are in this range, this ending; in that range, that ending. It branches cleanly and concludes cleanly. Systematically perfect.

But real relationships do not end cleanly. You broke up but sometimes think of them, you are together but anxious, you passed by but meet again 10 years later. "The end" does not exist in relationships. What exists is only "the state so far."

So I want to design this visual novel's endings not as "conclusions" but as "snapshots of the current state." The game ends, but the relationship has not ended. After the "Connected" ending, these two people can still fight, and after the "Late Love" ending, they can still meet again by chance. What the game shows is not the end of a story but one cross-section of it.

This perspective changes how endings are directed. Instead of a clean wrap-up, leaving an aftertaste. Making the player carry the question "So what happened after that?" Living with that question is itself the final emotion this game creates.

The most important experience this system creates is this. "Accepting the result of the choices I made." A good visual novel is not a game about finding the right answer, but a game about facing your own choices. And the best visual novel is a game where the relationship does not end even when the game does.

In the next part, we weave all the designs so far into a single structural diagram.

---

Next: The Complete Scenario Architecture — Weaving It All Together