Template Versions

Older cards can stay on Version 1.0. New cards should usually use Version 2.0, which asks for more specific story and craft information. Switching templates preserves matching fields where possible, but V2 has fields that V1 does not. If you edit in V2 and save, treat V2 as the current version of that card.

• Version 1.0: the older compact card format. It still works for existing personas.
• Version 2.0: the preferred format for new cards, richer autocomplete, and better semantic compression.
• Draft Guidance: temporary instructions for Generate Editable Draft and Auto-Complete. It helps with this generation only and is not permanent persona content.

What A Character Persona Controls

• Archive fields control discovery: visibility, category, fandom, tags, tagline, archive summary, and creator username.
• Tool fields help generation and source handling: draft guidance and canon reference notes.
• Identity fields anchor stable facts: name, age, gender, pronouns, setting, and appearance.
• Behavior fields define how the character acts under pressure: personality, speech style, contradictions, trust, stress, defenses, avoidance, wants, humor, intimacy, and failure patterns.
• Continuity fields define what must stay true: relationship to user, knowledge and memory state, shared/private history, hidden truths, boundaries, behavior rules, and development intent.
• Opening fields give the chat its start: system prompt, primary greeting, and additional greetings.

Version 2.0 Field Guide

• Visibility: whether the persona is private or can appear in the public archive.
• Category, fandom, and tags: where the card belongs in browsing, filters, and directories.
• Tagline and archive summary: public-facing hooks that explain the appeal quickly without dumping the whole card.
• Canon Reference Notes: stable source facts, vocabulary, continuity limits, body/species rules, or canon overrides.
• Setting: where and when this version belongs: canon era, AU, occupation, social world, or current situation.
• Speech Style: how the character sounds: rhythm, accent, language, vocabulary, formality, humor, evasions, and verbal tells.
• Relationship to User: the current or usual bond, plus the emotional terms of that bond.
• Core Contradictions: tensions that make the character alive, such as wanting closeness while resisting vulnerability.
• Trust & Attachment Pattern: what earns trust and what makes them test, cling, soften, pull away, or doubt.
• Conflict & Stress Response: what they do under pressure: freeze, deflect, overexplain, attack, appease, withdraw, joke, or control.
• Knowledge & Memory State: what they know, do not know, remember, misunderstand, or must not know in-scene.
• Shared and Private History: old events, routines, wounds, inside jokes, sore subjects, or established intimacy.
• Behavioral Tells: visible habits that reveal internal state without explaining it directly.
• Secrets & Hidden Truths: author/reader-known truths for tension and reveal pacing.
• Emotional Defenses: what they do instead of being honest: sarcasm, competence, caretaking, silence, anger, charm.
• What They Avoid: topics, feelings, admissions, dependencies, or relationship moves they resist.
• Private Wants: what they want but may not say directly.
• Humor & Play: how they manufacture humor, not just whether they are funny.
• Affection & Intimacy Style: how closeness, care, tenderness, desire, comfort, and touch appear when licensed.
• Attention Lens: what they notice first and what they miss. This is often voice-defining.
• Ordinary Texture: daily habits, rooms, food, rituals, objects, sensory details, and repeated human things.
• Boundaries / Do Not Assume: what the model must not decide, narrate, reveal, force, sexualize, or override.
• Behavior Rules: absolute habits and nevers, such as "does not lie" or "does not trust instantly."
• Failure & Doubt Pattern: what they fear they will fail at and how shame, guilt, or insecurity changes behavior.
• Development Intent: how the character is meant to grow over a long story without forcing it too quickly.
• System Prompt: extra high-priority writing instructions for this character's chat behavior.
• Primary and Additional Greetings: main and alternate openings for different starts.

Generate And Auto-Complete

Generate Editable Draft creates a full editable card from the fields and draft guidance you provide. Auto-Complete fills only blank fields from the information already present. Both tools should be treated as assistants, not final authority.

• Use Draft Guidance for AU premise, expected user role, relationship setup, tone, timeline, or greeting direction.
• Review generated fields before saving, especially boundaries, hidden truths, relationship setup, and greetings.
• Remove repeated ideas. The compiler will compress meaning, but a cleaner source card still helps.
• Do not leave generic filler just because it sounds polished.

Public And Private Personas

Private personas stay tied to your account. Public personas can be found in the archive, liked, reported, and copied when duplication is allowed. Public cards should be safe for other testers to read and use. Keep private real-world details, API keys, passwords, and personal records out of public cards.

Saving Across Card Types

Character and user persona cards are now more parallel. When the app offers a "save as user persona" or "save as character persona" path, it creates an editable draft in the other card type. Review the draft before saving, because relationship direction, greetings, and boundaries may need different wording after conversion.

Avoid These Problems

• Repeating the same idea in many fields instead of stating it clearly once.
• Using only adjectives without explaining how the behavior is made.
• Putting broad roleplay style instructions in the character card when they belong in Roleplay Style.
• Making the character instantly trust, love, forgive, or understand the user unless that is the premise.
• Writing greetings that decide the user's emotions, actions, consent, or dialogue.
• Leaving generated text unreviewed because it is long.

Need More Help?

If a persona behaves strangely in chat, use feedback or flag the response from inside the app. Include what felt wrong and what you expected the character to understand.