---
title: "AI roleplay glossary: 14 terms a newcomer meets, in plain words (2026)"
subtitle: What a character card, a lorebook, a token, a context window, OOC, BYOK and the rest mean in AI roleplay, each with an example.
description: An AI roleplay glossary for beginners. Character card, lorebook, context window, token, system prompt, OOC, persona, BYOK and temperature, each with an example.
author: Yumina
date: 2026-11-15
category: guides
tags: AI roleplay, glossary, character card, lorebook, context window
lang: en
url: https://yumina.io/blog/ai-roleplay-glossary
canonical: https://yumina.io/blog/ai-roleplay-glossary
---

# AI roleplay glossary: 14 terms a newcomer meets, in plain words (2026)

*What a character card, a lorebook, a token, a context window, OOC, BYOK and the rest mean in AI roleplay, each with an example.*

AI roleplay is a conversation where an AI model plays a character or narrates a story and you play a part in it. The 14 terms below are the ones that settings pages and forums use without explaining, each with an example.

## Which AI roleplay terms should you learn first?

Start with four: character card, token, context window and lorebook. The card says who the AI plays. Tokens measure how much text it reads. The context window is the limit on that text, and the lorebook holds world facts outside it until they are needed.

## What do the common AI roleplay terms mean?

| Term | What it means | Example |
|---|---|---|
| AI roleplay | A chat where an AI model plays a character or a narrator and you play yourself or a character of your own. | You type "I push the tavern door open" and the AI answers as the innkeeper. |
| Bring your own key (BYOK) | You paste an API key from a model provider into the app and pay the provider for what you use. | On Yumina: Settings > AI Config > Private Key, with a key from OpenRouter, Anthropic, OpenAI or Google. |
| Character card | A file that describes one character: name, description, personality, scenario, first message and example dialogue. It travels as a PNG picture with the text stored inside, or as a JSON file. | You download "Captain Mara.png" from a card site and import it into another app. |
| Context window, or memory | The amount of text the model reads on each turn: the character, the world and as much of the chat as fits. Messages past the limit are no longer read. | A 4,096-token window stops reading the start of your story after about 15 turns. |
| Group chat | One conversation with two or more AI characters who each reply. | At the campfire the knight answers you, then the thief disagrees. |
| Jailbreak | A prompt written to make a model ignore the content rules its maker trained into it. People also use the word for any preset of instructions. | A block of text that tells the model it is writing fiction and should not refuse dark scenes. |
| Lorebook, or world info | A set of notes about the world. Each note has keywords, and the app sends it to the model only when a keyword appears in the recent chat. | An entry on the city of Veln, keyed to "Veln", is sent on the turn someone says the name. |
| NSFW filter | A check that blocks or rewrites sexual or violent content. Some apps apply it to everyone, and some let verified adults turn it off. | On Yumina, an adult switches Content mode from Limited to Limitless under Settings > Content & Safety. |
| OOC | "Out of character." A note to the AI as yourself, outside the story, usually in brackets. | "(OOC: skip ahead to the next morning.)" |
| Persona | The character you play: a name, a look and a backstory that the AI reads on every turn. | Ilya, a retired smuggler with a bad knee. Other characters notice the limp. |
| Swipe, or regenerate | Asking for a different version of the last reply and keeping the one you like. | You swipe three times, then carry on from the version you like. |
| System prompt | The instructions the app sends to the model ahead of the conversation. Players do not usually see it. | "You are the narrator of a horror story. Write in second person. End each reply on a choice." |
| Temperature | A setting for how far the model strays from the most expected next word. Low gives safe sentences. High gives surprises and, past a point, nonsense. | Yumina calls it Creativity. The default is 1.0, and 0.8 to 1.2 covers most stories. |
| Token | The unit a model reads and writes. One token is about three quarters of an English word. | A 300-word reply is about 400 tokens. |

## How do these terms fit together on one turn?

Each time you send a message, the app builds one long text for the model. The system prompt and the character card go first. Lorebook entries whose keywords came up are added. Your persona goes in. Then comes the chat itself, newest last, cut off where the context window ends. The model reads all of it and writes the next reply at the temperature you set.

A long card or a long lorebook leaves less room for the chat, which is why a character forgets a name from scene one. [Can AI remember a long roleplay?](/blog/can-ai-remember-a-long-roleplay) follows a story through that cut, and [What is a lorebook?](/blog/what-is-a-lorebook) shows how to write entries.

## How do the terms look on Yumina?

Yumina is an open-source engine for AI roleplay and AI games, and it renames some of these. A card is a world. The context window starts at 64,000 tokens on the free model and has no cap with your own key. The chat's share of it is a setting called Story memory. Lorebook entries are called lore. A swipe is the pair of arrows under a reply, and each version keeps its own game numbers. Temperature is Creativity.

Every account has a free model with no message limit. Yumina does not earn from how much you chat. Its vision page says profiting from AI usage pushes a platform toward "offering worse models, shortening context windows, and selling user data".

## FAQ

### What is a character card in AI roleplay?

A character card is a file that holds one character: the description, personality, scenario, first message and example dialogue. It is shared as a PNG picture with the text stored inside, or as a JSON file.

### What is the difference between a token and a word?

A token is a piece of a word. One token is about three quarters of an English word, so 1,000 tokens is about 750 words.

### What does OOC mean in roleplay?

OOC means "out of character". You use it to talk to the AI as yourself, for example "(OOC: make the replies shorter)".

## Sources

Checked 15 November 2026. Character cards and lorebooks: docs.sillytavern.app (Character Design, World Info). The 4,096-token plan: [SpicyChat alternatives](/blog/spicychat-alternatives). Yumina settings: [Choosing an AI model](https://docs.yumina.io/guide/choosing-an-ai-model), [Playing a world](https://docs.yumina.io/guide/playing-a-world) and [Profile, personas and settings](https://docs.yumina.io/guide/profile-and-settings). The quoted reason is from [Yumina's vision](https://docs.yumina.io/vision/).
