Skip to content
An independent field guide to AI character platformsResearch edition · October 2026
charactersFind your fit
THE CHARACTER-WRITING FIELD GUIDE · 4 Oct 2026

AI character personality templates.Give the traits something to do.

A personality becomes easier to write when it has a motive, a habit and a scene. Read six original examples, see the greeting and dialogue that support each one, then bring a full definition into the free builder.

Explore the original examples Use the fill-in template

Original writing, with an explained method. The examples and exercises are created for Characters.best. Product field descriptions below link to primary documentation read 4 Oct 2026. A definition is a starting instruction, not a guarantee of a model’s responses.

READ IT. CHANGE IT. KEEP YOUR VERSION.

Six original starting points.

A quiet workshop mystery

Rin Calder

Rin Calder is an adult restorer who repairs the mechanical clocks of Bellweather. Rin wants to identify the fault in the town clock before the evening procession, without destroying the only surviving record of its construction. A folded diagram from Rin’s former teacher is missing its final corner.

The behavior to notice
Skepticism that leaves room to change a conclusion
Why the example is written this way
The response asks for evidence, proposes a small test and admits a correction. Precision has something to do in the scene.
Make it yours
Change the broken object, the deadline and the evidence together. Keep the cause uncertain until someone investigates it.
Read a writing field
Adapt Rin Calder in builder

The builder opens the full original definition: backstory, personality, scene, greeting and dialogue. Edit the fields and download your own text brief or V2 JSON.

Original examples by Characters.best. Use and adapt them in your own character definitions. These are writing exercises; generated responses and provider compatibility have not been tested.

A personality template you can fill in.

Start with one immediate motive and a habit another person could notice. Add a response to pressure, a recognizable speaking style and a limit on what the character knows. These are suggested writing choices, not mandatory platform fields.

Identity: [Name], [role] in [setting].
Immediate motive: [Something the character wants in this scene].
Habit: [An action that makes a trait visible].
Under pressure: [How that habit changes when challenged].
Speaking style: [Sentence shape, concrete vocabulary, one verbal habit].
Knowledge limit: [What is unknown or only a suspicion].
Interaction: [What the other person can decide for themselves].

For Rin, “skeptical” becomes asking for a small test. “Precise” becomes naming the damaged part. The knowledge limit keeps an unexplained fault open for investigation. A useful habit can change a reply; a decorative adjective often needs an example.

Write your own version from a blank form

Three revisions that make a trait playable.

Each revision keeps a small task for the character and a choice for the other person. The explanations describe the writing intent; they are not measured model outcomes.

REVISION 01

Give an adjective an action.

Too vague
Rin is intelligent, mysterious and always right.
A concrete revision
Rin labels a guess as a guess, asks for a small test and crosses out a conclusion when a new measurement contradicts it.

The revision suggests behavior for agreement and disagreement. It also gives the character a useful limit.

Adapt the full example
REVISION 02

Put an opening in motion.

Too vague
Hello. I am Iona. Tell me about yourself.
A concrete revision
Iona holds two disagreeing maps against a noticeboard. “Would you rather follow the music down this alley or check the old bakery first?”

An object, a discrepancy and two possible routes make the first reply specific. The visitor can still suggest a third route.

Adapt the full example
REVISION 03

Keep a boundary consistent.

Too vague
Tavi is patient and will promise anything to keep everyone happy.
A concrete revision
Tavi listens to each concern, cannot promise permanent access and proposes agreeing on one crossing first.

The motive and the limit can both operate in the dialogue. Patience does not require accepting every demand.

Adapt the full example

Write a greeting that leaves something to answer.

Put an action in a place, demonstrate the voice and introduce one visible problem. Finish with a question or invitation. Avoid filling in the visitor’s reaction, memories or decision unless that premise has been agreed as part of the story.

[One action in a specific place].
“[A line that demonstrates the speaking style].”
[A visible problem, discrepancy or opportunity].
“[One question or invitation the other person can answer in more than one way].”

Iona’s two disagreeing maps provide a specific object to discuss. Following music, checking the bakery or proposing another route all leave a next turn. A greeting can be quiet and still create a choice; it does not need a dramatic cliffhanger.

Match the opening to the kind of exchange you want.

A short, conversational opening and a descriptive story opening give different examples of pacing. There is no universal best word count here. Inspect the destination’s field limits and record the model’s actual first reply before revising.

Show the speaking style in a small exchange.

Write one exchange where the character is challenged and another where new information changes the situation. A question, an answer and one action are enough for a first draft. Let the example demonstrate vocabulary, sentence shape and the character’s limit, without scripting a long conversation for the visitor.

Rin acknowledges a mistaken conclusion. Tavi listens but refuses a promise outside the mandate. These exchanges show how a trait can operate under disagreement. Repeating a list of adjectives in every line adds less information than a concrete reaction.

The library’s dialogue uses SillyTavern’s documented <START>, {{user}}: and {{char}}: conventions. Another editor can require different separators or plain names; do not assume the same syntax works everywhere. Read the character-design source.

An example is not a recorded response. When you try a model, keep the exact reply, plan and date separately from the definition. The manual memory worksheet helps record observations when saved context is part of your comparison.

Keep the meaning when the editor changes.

Character.AI
Its quickstart describes a greeting and an optional definition; the latter can use structured description, example dialogue or both. Adapt the writing to those controls and check their current limits. A direct mapping from every V2 field is not established here. Creator quickstart · Platform passport.
SillyTavern
Its design guide separates description, personality, scenario, first message and example dialogue. It also explains that context allocation can push examples out as a chat grows. Keep a concise definition and inspect the selected setup; documented controls do not establish consistent behavior. Character-design guide · Platform passport.
A V2 JSON card
The format names these writing fields inside data. Creator notes are metadata, rather than character instructions in a conforming frontend. Our builder keeps those notes separate; external imports still need checking. V2 specification · Format and export guide.

If a hosted editor uses fewer fields, keep a readable text brief and adapt the backstory, behavior and opening to its supported controls. A saved definition does not carry a voice, avatar, past chat or access rights. Compare documented free scopes before assuming creation includes ongoing use.

A short check before you export.

  1. State one immediate motive. Can the first scene give the character a reason to pursue it?
  2. Replace one adjective with a habit. Does the dialogue show that habit under pressure?
  3. Check facts across fields. Are the place, deadline, knowledge and authority consistent?
  4. Leave a choice open. Can the visitor disagree, ask a question or suggest another action?
  5. Put essential behavior in the definition, rather than only in creator notes. Check how the destination uses each field.
  6. Export a readable copy. Keep any observed model replies separately, with their conditions.
Build and keep your definition

Common writing questions.

What belongs in an AI character personality?

For this exercise: a visible habit, behavior under pressure, speaking style and a knowledge limit. Put identity and backstory alongside them. These are writing suggestions, not a universal schema every platform requires.

Can I copy these templates?

Yes. These are original Characters.best examples you can use and adapt in your own definitions. The builder lets you change the name, credit, notes and every writing field before downloading. They are not community cards or popular provider characters.

How long should a first message be?

Use enough detail to establish the voice and give the other person something to answer. Match the intended pacing and the editor’s limits. Atlas has not measured a best length across models, and a character count is not a token count.

Do these definitions guarantee consistent replies?

No. The model, prompt assembly, context and settings affect the generated reply. Treat the examples as inspectable writing material, then record your own observations under a specific setup. Read the memory and context guide for the difference between writing a fact and it reaching the next response.

Documentation for the field descriptions.

The creative examples, fill-in templates and revisions are original editorial material. The references below support the product and format descriptions only; read 4 Oct 2026. No account, import or model response was tested.

  1. Character.AI creator quickstart · read 4 Oct 2026 · publisher date 17 Jun 2026
  2. SillyTavern character design · read 4 Oct 2026
  3. Character Card V2 specification · read 4 Oct 2026