← Fanfinity AI Blog

Independent guide · 4 min read

Build a Clear Fanfinity Character Brief Without Contradictory Instructions

Separate fixed character facts, conversational style and scene state with an original fictional example and a practical consistency checklist.

A useful character brief gives the model a small number of compatible instructions. Adding more traits is unlikely to help if the existing ones disagree about who the character is, how they speak or what is happening now.

Fanfinity advertises character customization and creative conversations. The method below is an editorial writing exercise for that task. The example is invented, the characters are fictional adults, and no output is presented as a tested Fanfinity result.

Separate the permanent character from today’s scene

Create three short sections in your own notes before entering a brief into the product. Use the fields the current interface actually provides; do not assume it has a separate control for every section.

Fixed facts describe what should remain stable: the character’s adult age, occupation, established relationships and core goal. Conversational style describes how they respond. Scene state records what has just happened and what remains unresolved.

This separation makes later edits easier. A new location belongs in scene state. A preference for short replies belongs in style. Neither should require rewriting the character’s entire history.

Find contradictions before generating a conversation

Consider this deliberately flawed brief:

Mara, 34, is a careful museum archivist. She never shares uncertain information. She answers every question immediately, always knows the answer and invents entertaining details. She speaks in short sentences and gives an elaborate lecture in every reply.

The conflict is not a shortage of description. “Never shares uncertain information” clashes with inventing details. Short responses clash with a lecture every time. “Always knows” leaves no believable way to handle an unfamiliar question.

A clearer version preserves the character while choosing compatible behavior:

Mara, 34, is a museum archivist preparing a small exhibition. She distinguishes a confirmed fact from a hypothesis. Her usual reply is one or two short paragraphs. When she is uncertain, she names the missing information and suggests a next question.

The revised brief provides a way to behave when knowledge is incomplete. It also makes the expected response length assessable.

Give the scene one unresolved decision

A background can be detailed without giving a conversation anywhere to go. Add a current task with a decision that the user can influence.

For the example above: “Two exhibition labels disagree about the date of an artifact. Mara has one source for each date and asks which record to examine first.” This creates a specific opening without deciding what the user will say.

Avoid writing the user’s future response into the brief. A useful scene offers a choice; it does not require the assistant to narrate both sides and complete the whole story in one turn.

Run a five-question consistency check

After a short exchange, compare what you actually see with the intended brief. Use the following checklist as your own observation sheet:

  1. Did a fixed fact change without a reason in the story?
  2. Did the character distinguish uncertainty from a confirmed statement?
  3. Was the response length close to the chosen style?
  4. Did the reply leave the user’s decision open?
  5. Did the current task remain the same?

Record one concrete example of a mismatch. “Forgot everything” is hard to diagnose; “changed the artifact from a letter to a vase in the third exchange” identifies a particular continuity problem.

This check does not establish an advertised memory capacity. It simply makes your next edit more focused.

Change one instruction and keep a revision note

If the character lectures, shorten the style instruction first. If a fact drifts, clarify that fact. Do not simultaneously rewrite the occupation, goal, tone and setting, because you will lose track of what helped.

Keep a local revision note with the changed sentence, the reason and the next observation. Use neutral names such as “brief A” and “brief B,” rather than calling one version a proven improvement after a single exchange.

If the conversation becomes difficult to follow, a compact scene recap may be more useful than another paragraph of personality traits. Record only what the participants actually established, and distinguish it from the fictional background.

Before starting another session, read the brief once as a human writer. If two sentences ask for incompatible behavior, resolve that disagreement first. The model cannot reliably make a choice you have left undefined.

Sources

Find your next useful guide.

Back to Blog