Glossary
Character cards, lorebooks, context windows, branching: the vocabulary of AI roleplay explained in plain English, with a note on how each one works on Reverie.
01
The building blocks of a character: what it is made of, how it travels between platforms and how you show up in the scene.
A portable file that holds a character: name, description, personality, scenario, example dialogue and greeting. The common interchange format is the Tavern V2 card, a PNG with the definition embedded as JSON (or a plain JSON file). Reverie's character importer reads SillyTavern and CharacterHub cards directly.
The written instructions that tell the AI who a character is: personality, backstory, speech patterns, motivations and limits. A strong definition shows rather than tells, so the model can stay consistent for hundreds of messages. On Reverie it is written in the character creator, or generated conversationally and refined afterwards.
The opening message a character sends when a new chat starts. It sets the scene, the point of view and the tone before you type a word, which is why experienced creators spend as much time on the greeting as on the personality.
Short sample exchanges included in a character definition to show the model how the character actually talks: sentence length, slang, humour, how it handles conflict. Models imitate examples far more reliably than adjectives.
Who you are in the conversation: your name, appearance, background and how characters should treat you. Reverie supports multiple personas per account, and a character keeps a separate history with each one, so the shy student and the war veteran you play never blur together.
See alsoPersonas on Reverie
An opt-in setting for verified adults that allows explicit content in chats and reveals characters tagged as adult. It is off by default, requires signing in and confirming your age, and never overrides the community guidelines.
See alsoNSFW AI chatCommunity guidelines
Labels creators attach to a character (genre, setting, dynamic, fandom) so other people can find it. Reverie turns popular tags into browsable hashtag pages that group characters across languages.
See alsoBrowse hashtags
A tool that brings characters and conversations from other platforms into a new one without rebuilding them. Reverie imports characters from SillyTavern and CharacterHub cards, plain text and screenshots, and chat history from Character.AI, SillyTavern, RisuAI, ChatGPT and Claude exports.
A user who publishes characters, stories or plugins for others to use. On Reverie, creators get analytics on how their characters perform and a revenue share when those characters are popular.
See alsoFor creators
02
How a character sees the conversation, what it can hold at once and what it keeps between sessions.
The maximum amount of text a model can consider at once, measured in tokens. Everything the character 'sees' on a turn (its definition, your persona, World Book entries, memory and the recent messages) has to fit inside it. Bigger windows cost more per message, which is why memory systems exist.
See alsoAI models
The unit a language model reads and writes, roughly three-quarters of an English word (less for other scripts). Context windows are measured in tokens, and paid usage on Reverie is metered in credits that scale with tokens used.
See alsoPricing and credits
Durable facts about you and the story (names, preferences, milestones, what changed) stored outside the context window and brought back in later sessions. On Reverie memory is visible and editable: you can review, correct, pin or delete any entry.
See alsoAI chat with memoryMemories
As a conversation grows past the context window, older scenes are condensed into a running narrative that preserves plot state, emotional dynamics and unresolved threads. Recent messages stay verbatim; the summary carries the rest.
See alsoHow memory works
A memory entry marked so that later conversations cannot rewrite it. Use it for facts that must never drift: a character's canonical age, an agreement the story hinges on, a rule of the world.
A collection of keyword-triggered entries about places, factions, rules, history and supporting cast. An entry is injected into the context only when the scene makes it relevant, so lore stays consistent across every chat and character that uses the book. SillyTavern calls the same idea a lorebook.
See alsoWorld BooksMemory and lore guide
The hidden instructions sent to the model before your message on every turn. On a character platform it is assembled automatically from the character definition, your persona, active World Book entries and memory; you shape it through those pieces rather than editing it directly.
03
The controls you use while a scene is running, from steering a single reply to running a whole cast.
Collaborative fiction improvised in turns between you and one or more AI characters. Unlike a chatbot that answers questions, a roleplay character stays in the scene, keeps its motives and reacts to what your persona does.
See alsoAI roleplayRoleplay ideas
A conversation with several characters in the same scene. Reverie's orchestration decides who speaks next based on the flow of the conversation, so characters interrupt, answer each other and stay in their own voices instead of taking rigid turns.
See alsoAI group chat
Splitting a chat at any earlier message to explore a different outcome. The original thread stays intact, so a bad decision or a plot twist you want to try never costs you the version you liked.
See alsoChat branching
Asking for a different version of the character's last reply without changing your own message. SillyTavern users know it as a swipe; it is the fastest fix when a reply drifts out of character.
Short commands typed into the message box (starting with a slash) that steer one turn: pacing, point of view, length, tone or a direct instruction to the narrator, without rewriting the character definition.
See alsoSlash commands
A note to the model rather than to the character, usually written in parentheses or brackets, such as asking to slow the scene down or skip ahead. The character does not treat it as dialogue.
Messages a character sends first, after a pause, based on the relationship and recent events rather than a timer. The feature behind 'AI that texts first'.
See alsoAI that texts first
Posts characters publish on their own (thoughts, updates, images) in a social-style feed. You can react, comment or start a conversation from any moment.
See alsoMoments feed
Talking to a character out loud: your speech is transcribed, the character replies in a synthesised voice, and its personality, persona context and memory stay available during the call. Voice cloning lets a creator design or reproduce a specific voice.
See alsoAI voice callsVoice cloning
Add-ons that change what a conversation can do: prompt plugins that tune behaviour, tool plugins such as RPG status panels, inventories and trackers, and interactive HTML panels that render inside the chat.
See alsoRoleplay plugins
04
The formats that turn a chat into something you can replay, direct or read as a book.
A chaptered, branching narrative where your choices change later scenes. Characters carry relationships and progression from chapter to chapter, so a story can be replayed for a different outcome.
See alsoInteractive stories
Long-form prose writing with an AI co-author. You direct the narrative and edit, delete or regenerate individual segments rather than trading chat messages.
See alsoAI novel writer
An instruction to the narrator in Novel mode (pace, tone, point of view, what happens next) that shapes the next segment without appearing in the text itself.
Converting a roleplay transcript into narrative prose while preserving the dialogue, events and chronology. Useful for turning a great session into a readable, editable story.
See alsoChat to novel
A tool that produces a premise, a scene or a whole outline from a short prompt. On Reverie it seeds interactive stories and Novel mode drafts that you then play or edit.
See alsoAI story generator
05
What is generating the words, how it is paid for and the workarounds you don't need here.
The AI model that writes a character's replies. Different models have different strengths in prose, memory handling, speed and how strict they are. Reverie offers several (Gemini, DeepSeek, GLM and more) and lets you switch inside a chat.
See alsoAI models
A per-model option that lets the AI think through a reply before writing it, with the thinking shown in a collapsible pane. Helpful for complex scenes and puzzles; slower and costlier for casual chat.
See alsoAI models
Reverie's usage currency for premium models and for image and video generation. Purchased credits never expire, subscriptions include a monthly allowance, and a free model is always available with fair-use rate limits instead of a daily quota.
See alsoPricingUnlimited chat
A prompt trick used to push a chatbot past its content filter. On Reverie there is no added filter to defeat: adult content is an account setting for verified adults, and the remaining limits are the community guidelines.
See alsoNo filter AI chat
A relay that forwards a chat site's requests to a model provider using someone's API key, common on platforms that ship without models of their own. It adds setup, instability and a dependency on keys that can be revoked. Reverie includes its models, so none of that applies.
See alsoReverie vs Janitor AI
A website that installs like a native app, with an icon, offline fallback and shared sessions across devices. Reverie runs as a PWA on phones and desktops, alongside its Discord and Telegram bots.
See alsoDiscord botTelegram bot
Put the words to work
Every term above is a feature you can try in a couple of clicks. Free to start, with no daily message quota on the free model.