Search “Space”
Sign in
Lyra University
›
Personas & guidelines

Personas & guidelines

Personas decide who you are in the scene; guidelines decide how the scene plays. Together they let you set the stage without rewriting it every time.

Personas — write as a character

A persona is the voice you speak in. Instead of replying as a generic “you,” you can step in as a named character with their own background and manner, and the cast will respond to that person. Set a persona once and it carries across the chat.
·
A persona has a name, an optional title (a role or relationship), and a description of who you are.
·
Pick a persona per chat from the chat settings — switch any time.
·
Great for recurring characters you like to play, or for matching a specific story’s tone.

Create one with the assistant

You don’t have to write a persona from scratch. Open the persona creator (You → Personas → New persona) and tell the Persona Assistant who you want to be — it drafts the name, title, and description for you, and you can ask it to refine any of them.
1
Open a new persona and describe the character you want to play — a line is enough ("a jaded war correspondent in her 40s, gruff but secretly kind").
2
The assistant fills in the fields. Ask it to adjust tone, age, or backstory until it fits.
3
Tweak anything by hand and set it as your default if you like — changes save as you make them.
The Persona Assistant (right) drafts your name, title, and description from a one-line idea — refine by chatting.

Guidelines — steer tone and content

Guidelines are free-form notes that tell the model how you want the scene to go — pacing, mood, what to lean into, what to avoid. Think “keep replies short and punchy,” or “slow-burn, lots of inner monologue.” They apply to the whole chat and you can edit them at any point.
Guidelines are about style and direction, not world facts. Put who-and-what details in the story’s definition or worldbook; put how-it-should-feel in guidelines.

Reasoning

Turn on reasoning when a scene is intricate and you want the model to think before it writes — weighing motivations and continuity first, then composing the reply. It is off by default to keep responses fast; flip it on per chat when the moment calls for more deliberation.