AI roleplay terms explained: OOC, persona, context, and more
Updated
This glossary explains common terms in AI roleplay and character chat. Many come from tabletop games and forum roleplay, and others come from how AI models work.
Playing a story
- Roleplay (RP): Telling a story together by each playing a part. In AI roleplay, you play your character and the AI plays the others.
- In character (IC): Writing as your character, inside the story.
- Out of character (OOC): Stepping outside the story to say something as yourself, usually in brackets or parentheses: “(OOC: let’s skip ahead to the morning.)” See how to steer an AI story.
- Persona: The character you play: your name, appearance, and personality in the story, as opposed to the AI’s characters.
- NPC (non-player character): A character controlled by the AI or the story rather than by you, such as a shopkeeper, a guard, or a rival. The term comes from tabletop and video games.
- Scenario: The setup of a story: where and when it takes place, and the situation at the start.
- Greeting / first message / opening: The message that starts a story. It strongly shapes the length and style of what follows. See how to write a great opening scene.
- Time skip: Jumping forward in the story with a line of narration, such as “Three days later…”.
- Slow burn: A story where something, often a relationship or a mystery, develops gradually over many scenes.
- Godmodding: Controlling another character’s actions or making your own character impossible to beat. Considered bad form, because it leaves nothing for the other side to do.
Formatting
- Asterisks (*like this*): The common convention for actions and narration: she slides the letter across the table.
- Quotation marks (“like this”): Spoken dialogue.
- Point of view (POV): Whose perspective the story is told from. First person (“I open the door”), second person (“you open the door”), or third person (“she opens the door”).
Characters and memory
- Character definition: The description of an AI character (personality, backstory, voice, example lines) that is sent to the model with every reply. See how to write a great AI character.
- Example dialogue: Sample lines that show how a character talks. A reliable way to set a voice. See how to give a character a distinct voice.
- Context window: The amount of text an AI model can read at once. In a long story, older messages eventually no longer fit.
- Token: The unit AI models measure text in, roughly a word or part of a word. Context windows are measured in tokens.
- Memory / long-term memory: Ways an app keeps important information available after it falls out of the context window, for example by summarizing or saving key facts. See why AI characters forget.
- Pinned memory: A fact or message you mark so the app keeps it available, in apps that offer this.
- Lorebook / world info: A collection of entries about a story’s world, such as places, factions, and history, that some apps add to the context when they become relevant.
- Character drift: When a character slowly loses their personality or voice over a long conversation, often sliding toward a generic assistant tone. See how to keep AI characters consistent.
Replies and controls
- Swipe / regenerate / reroll: Asking for a different version of the AI’s last reply.
- Editing a reply: Changing the AI’s reply directly, often to fix a mistake before it becomes part of the story.
- Repetition loop: When replies keep reusing the same phrases or structure. See why AI replies get repetitive.
- Multi-character scene / group chat: A scene with several AI characters at once, who react to you and to each other. See multi-character scenes.
- Model / LLM: The large language model that generates the characters’ replies.