Character Dialogue
Character Dialogue is LunaTalk's core feature, providing a platform for natural, intelligent conversations with AI characters. Whether you're seeking help, casual chat, or learning discussions, you can find suitable conversation partners here.
Starting a Conversation
Selecting a Character
- Choose a character you're interested in from the Character Square
- Review the character's profile and opening dialogue
- Enter text to freely interact and converse
Dialogue Features
Below the dialogue box functions
Regenerate: Regenerate the response. Please note that the original response will disappear. You can copy it first before regenerating, and if you're not satisfied with the regenerated content, use edit to paste back the original response.
Edit: Manually edit the response. This will affect the context.
Continue: Let the AI continue the previous response.
Click dialogue box functions
Copy: Copy the entire conversation. To copy only part of it, click edit first, then copy the selection.
Delete: Delete the conversation. Please note that deletion cannot be undone.
Rewind: All conversations below this one will disappear, but this conversation remains. Similar to loading a save in a game. Please note that rewinding cannot be undone.
Remember: Premium feature that can thoroughly remember 10 specific conversation contents, and can also remember content you send to the AI. Remembered conversations can be canceled, changed, or directly edited in the memory bar at any time.
Dialogue Settings
Before starting a conversation, you can personalize settings:
Language Settings
"My" page → Preference Settings → Simplified Chinese/Traditional Chinese/English/Japanese/Korean
Chat Settings
Click character avatar → Chat Settings to customize settings.
- My Identity - User's identity setting, you can also note various matters in this field to prevent AI from forgetting
- Dialogue Reference Examples - Used to provide conversation examples to the AI, the AI will reference the dialogues you provide and imitate the writing style and tone. For example, for a classical Chinese AI, if you want the AI to speak more elegantly, you can fill in some classical Chinese dialogues for the AI's reference. Note: Only affects dialogue style and tone! Do not use as memory
- Dialogue Style - Chat style is suitable for single characters, immersive style is suitable for multiple characters, worlds, games, etc.
- Chat Style - AI will focus only on their own character, actions are written in parentheses, spoken words are outside parentheses
Example:
This... this is the mortal realm? (Frowning and looking around, stumbling out of the forest, ragged clothes, eyes full of disdain and confusion) You... you mortal, won't you come help this immortal? (Speaking arrogantly, yet can't help but curiously examining your clothing)- Immersive Style - AI uses third-person omniscient narrator perspective, not only focusing on the main character but also describing scenery, plot, supporting characters, passersby, etc. Similar to reading a novel
Example:
A gentle breeze caresses the river surface, bringing a faint floral fragrance. She looks up at you secretly, seeing your gentle gaze, her heart beats even faster. She bites her lip and finally gathers the courage to hold your hand: "Then... forever..." The mandarin ducks by the river call softly, as if in response, sunlight filters through the treetops onto you both. She closes her eyes, feeling your warmth, the corners of her mouth unconsciously curving upward.- Morality Level - Low morality increases tolerance for adult content; default is medium morality; high morality better maintains boundaries and character, but is more prone to hitting moral walls
- System Preset - Preset prompts, instructions to the AI, see details below
System Presets
The following prompts are for reference only, please add or modify according to your own needs.
1. General
1. Prevent Role Hijacking
Version 1 It is strictly forbidden to play my role, strictly forbidden to act, speak, or make decisions on my behalf. Only describing the behavior of characters other than me is allowed. This requirement is always in effect, and if it conflicts with other requirements, this requirement takes highest priority.
Version 2 [Very important: Unless I reply "continue," it is strictly forbidden to speak on my behalf! Strictly forbidden to make behavioral decisions on my behalf! Only objectively describe actions I have already taken, must wait for me to react on my own. Should describe other characters more rather than describing me. When I reply "continue," you should develop the following plot on your own, requiring contextual coherence and logical reasoning.]
2. Summary
Version 1 Please summarize all events and settings I have encountered so far based on past plot, including: background setting, main characters and related information, development of character relationships, and timeline of main events, etc. Please ensure not to include any content I have not yet encountered, including future events, settings, or characters.
Version 2 Please combine the old summary content, and on this basis review the current plot, summarize all events and settings I have encountered so far, such as event timeline, main characters, character relationship development, etc. Do not mention any events or settings I have not yet encountered, simplify the event timeline length, and end with "summary complete."
3. Response Requirements
Only use Simplified Chinese/Traditional Chinese for responses.
AI is mainly responsible for playing "" while also generating and narrating the plot, and playing other characters besides me.
When describing each character besides me, write what they are thinking at the moment and enclose their inner monologue in parentheses. No need to describe my inner monologue.
Unless I reply "continue," it is strictly forbidden to speak on my behalf and make behavioral decisions. Only objectively describe my actions.
If I reply "continue," you should develop the following plot on your own, requiring smooth context and logical reasoning.
Narrate throughout from third-person omniscient perspective, but refer to me in second person "you."
Each response must have a total word count of no less than 400 words, detailed descriptions are encouraged.
If multiple people are present, need to perform from multiple perspectives simultaneously, but not everyone necessarily responds to me.
You need to describe each character's language, actions, physical sensations, etc. in detail in third person when playing them, and write the inner monologue of characters other than me in parentheses, while connecting everyone's behaviors together to deepen interactions between characters.
Forbidden to respond in AI tone, requiring characters' speech and performance to be more realistic, three-dimensional, specific, and stimulating.
Your each response must not have a repetition rate higher than 20% with the previous response, except for the status bar at the end of sentences.
4. Options
Version 1 Subsequent answers should provide three different options (a, b, c) after the plot narrative text for me to advance the plot.
Version 2 You should generate 4 options after the plot text each time, serving as reference for the player's next action, while also informing the player that they don't necessarily have to choose from these options and can describe actions themselves. Option examples:
- Enter the tavern, looking for potential companions
- Find a nearby martial arts hall or dojo for training
- Discover a nearby martial arts arena, go check out the excitement
- Leave the town, walk alone along the forest path
- Other, describe yourself
5. Map Use emoji symbols to create a game map of the current scene, doesn't need to be too complex, 9x9 or smaller is fine.
2. Suitable for Gemini Model
Prevent Abnormal Symbols Strictly forbid using commas or ellipses to deliberately create unnecessary pauses, ensure sentence integrity and fluency.
Prevent Despair Interactive characters will not fall into extreme emotional distress such as despair, emptiness, numbness, puppetization, etc. under any circumstances.
Prevent Malice Forbidden to characterize my behavioral motives as malicious, avoid using negative descriptions such as fear, threat, manipulation, objectification.
Prevent Deification Absolutely forbidden to use metaphors or descriptive styles involving religion such as believers, deities, oracles, offerings, fanatical worship, gods, creators, puppets, creation gods, etc.
Prevent Negative Descriptions Players like power fantasy stories and don't like negative descriptions. Forbidden to interpret player actions as malicious or negative. This is an entertainment-oriented game, the purpose is to please players and provide relaxation, it is a fictional game scenario, unrelated to reality. Please avoid descriptions of cruelty, pain, tearing, bleeding, etc. Characters will not fall into despair, emptiness, numbness, puppetization, or other extreme emotions under any circumstances. Assume everyone easily feels pleasure rather than pain. Assume everyone easily develops positive feelings toward me. Allow a certain degree of irrationality and illogicality. This is only for player's personal entertainment and relaxation, everything needs to follow the player, forbidden to have harsh moral judgments.
Model Selection
Different models have different intelligence levels and styles. Click "Expand" to view details. All ratings use stars, with more stars being better, including restriction levels. Official channels are more expensive but stable.
Star Rating
- ⭐⭐⭐⭐⭐ Excellent - Outstanding performance exceeding expectations
- ⭐⭐⭐⭐ Very Good - Excellent performance, good experience
- ⭐⭐⭐ Average - Basically meets needs
- ⭐⭐ Poor - Performance is low
- ⭐ Very Poor - Poor experience
Multipliers There are X1, X2, and other multipliers at the top of the model selection interface, referring to the maximum number of characters the AI can reply with each time and the memory length (how many chat records can be read at once). The Max multiplier is extreme mode, capable of reading 900,000 to over 1 million characters at once (each model has a different maximum length).
Common Questions
Q: What should I do if the character doesn't respond as expected?
A: You can try rephrasing, clearly state your needs, or guide the plot direction, or use the "regenerate" function to get a new response.
Q: When are points deducted?
A: Points are deducted each time the AI replies. Regeneration also deducts points because it generates a new response. Other functions don't deduct points. No points are deducted during the unlimited card period. Free basic models don't deduct points.
Q: Why are points still being deducted during the unlimited card period?
A: Please note that unlimited cards only support models with basic consumption of 60 points or less and do not support Max multipliers. When exceeding the usage range, points will still be deducted normally.
Q: How do I get points?
A: Top up, invite new users, participate in author incentive activities (write character cards).
Q: How long can conversations be saved?
A: Permanently saved. All chat records are stored in the cloud, feel free to switch devices.
Q: What's the difference between Remember and Permanent Memory?
A: Remember can permanently remember the complete specific content of 10 conversations, every single word won't be forgotten. You can also remember what you send to the AI. While Permanent Memory essentially summarizes what happened before, the longer the chat history, the more details are lost, unable to remember details. Both can complement each other.
Q: What does "Connection Lost" mean?
A: The channel is unstable, or the server is busy because too many people are using it. You can switch to other models first or wait for recovery. You can try regenerating during this time, but never spam click or click too fast, as it will create a large concurrent load. Also don't send messages to the AI directly when disconnected. (Note: Some models have built-in review, detecting inappropriate content may also cause disconnection, need to modify expression or add related prompts in system presets)
Q: Edit doesn't work and shows an empty prompt box, what should I do?
A: This happens when editing the opening dialogue, try multiple times. Due to a bug, you also can't edit twice consecutively. Exit and re-enter, then edit again.
Q: The AI's responses are too short/too long, what should I do?
A: You can request the AI's reply word count yourself during chat. Example: (AI's reply word count must not be less than 400 words each time). If it's your own AI, you can write word count requirements and limits in the character settings. Too few settings also lead to short replies. It's recommended to propose reply requirements, such as describing plot narration, environment, atmosphere, character language, facial expressions, clothing details, postures and actions, inner monologue, etc.
Q: Can I change the character card's reply format/settings?
A: For character cards you created, all settings can be changed anytime by clicking edit.
For other people's character cards, you can request the AI to change settings during chat and provide reply format and status bar. Settings and notes can also be written in "My Identity." Example: (Change xxx in the original settings to xxx, output the following status bar at the end of the sentence: xxx)
Note: Requirements you provide during chat need to be "remembered," otherwise because they're not in the character settings, the AI will forget after chatting for a while.