Why Flat Characters Destroy Otherwise Good Stories
A flat character is one the reader can summarize in a single adjective: brave, greedy, kind. Memorable characters resist single-adjective summaries because they contain contradictions — they are brave about some things and cowardly about others, generous to strangers but cruel to people they love. These contradictions are not random; they come from the character's wound. Something happened to this person that shaped a core belief about the world, and that belief drives every decision they make — including the ones that hurt them. AI is excellent at generating psychologically coherent character profiles from a wound backwards because it can draw on patterns from thousands of stories and identify the specific behaviors a particular wound would produce.
The Want, Need, and Flaw Triangle
Every compelling character is defined by three elements that exist in tension: what they want (their stated goal), what they need (the emotional truth they are avoiding), and their flaw (the behavior that prevents them from getting what they need). The character pursues the want obsessively while their flaw blocks the need. The story resolves when they either achieve the need by surrendering the flaw — or fail to, which is tragedy. AI can generate dozens of want-need-flaw combinations from a basic character premise, letting you find the specific triangle that creates the most interesting dramatic conflict for your plot. Specifying your genre and story arc helps AI align the triangle with what your narrative actually needs from the character.
How to Give a Character a Distinctive Voice
Voice is the sum of word choice, sentence length, what a character says, and crucially what they never say. Two characters who share the same information will deliver it completely differently depending on their education, class background, emotional state, and relationship to the listener. AI can generate sample dialogue that demonstrates voice once you define the parameters: education level, cultural background, emotional guardedness, characteristic speech patterns, and what topics this person avoids at all costs. Providing five lines of dialogue for AI to analyze and replicate is more effective than describing the voice abstractly — it gives the model a concrete register to match.