AI companion apps come with a vocabulary of their own. A page that promises "unlimited memory" may mean a bigger context window, an editable memory list or an automatic summary. A page that says "uncensored" may mean one of several filters is relaxed. This glossary explains each term in a few lines and then shows how named apps such as Character.AI, SpicyChat, Janitor AI, CrushOn.AI, Kindroid and Nomi put it to use.
Definitions follow the sources we trust most for each term: vendor documentation for how language models work, the laws and regulators for legal definitions, and the published specifications for file formats. Where a term is community slang with no formal definition, such as "roleplay AI" or "character prompt", we say so. App facts come from official help centers, wikis, policies and store listings checked in October 2026. We did not create accounts or test the apps.
The entries fall into three groups: what the product is (AI companion, roleplay AI), how a character is defined and remembered (character prompt, character card, context window, long-term memory) and the rules around content and voice (NSFW filter, voice cloning).
New to the topic? Start with what an AI companion is, then read how AI girlfriend apps work. If you want to compare services rather than learn the words, go to the ranking of every app or how to choose an AI companion app.
Terms at a Glance
- AI companion
- A chatbot designed to sustain an ongoing, human-like relationship with a user, typically by remembering earlier chats and talking about personal matters.
- Roleplay AI
- A chatbot used to act out characters, scenarios and stories in conversation rather than to answer questions as an assistant.
- Character prompt
- The standing instructions that tell a chatbot who it is playing: personality, backstory, speaking style and scenario.
- Character card
- A JSON file, often embedded in a PNG image, that defines a roleplay character; the community has published V1, V2 and V3 versions of the format.
- Context window
- The amount of text, counted in tokens, that a language model can take into account when it writes a reply, including the reply itself.
- Long-term memory
- Information an app stores about you or a conversation beyond a single chat and brings back when relevant; each app implements it differently.
- NSFW filter
- A moderation layer that classifies or blocks sexually explicit material in what a user sends or what the model returns.
- Voice cloning
- Synthesizing speech that imitates a specific person's voice from a sample; consent rules come from providers and laws.
All Glossary Pages
Sources
Official pages and store listings we read for this page (last check October 20, 2026).