To the left of the chat room, you'll find tools that make your conversation richer. Learning them one by one will let you enjoy things much more comfortably.
This is the menu you use when a conversation has gotten too long, or when you want to reset the mood. A new chat room starts a fresh conversation with the same work, right from the beginning.
When a huge amount of conversation piles up in one chat room, the AI can lose track of context. Grill Swap is used to refresh the flow and lay down a new round.When you use it, settings like long-term memory, the most recent 10 chats, user notes, persona, and guidance all carry over together into the new chat room, so you can continue the story in the new round.
This is where you directly set the length and depth of replies.
Output Length: Sets the approximate length of a single reply. The default is 2,000 characters. (However, due to the nature of AI and the flow of the story, please think of this as a reference value rather than an exact character count.)
Reasoning Length: Sets how much the AI thinks internally before creating a reply. You can choose your preferred reasoning length/value from the reasoning values set for each model — the higher it is, the deeper and longer the AI thinks, but the cost increases.
Streaming Mode (on by default): As the AI outputs sentences, they appear immediately in order; if disabled, instead of displaying the reply as it's completed in real time, the entire completed reply is shown all at once after it's finished.
Show AI Reasoning Process: If you want to see the process the AI goes through while creating its answer, turning this on lets you check the reasoning process.
AI Model Switching Macro: Alternates between multiple models in order when sending regular chats.
User Guidance: Write down instructions the AI should reference in the current conversation — such as guidelines or tone requests — and they'll be reflected in every reply.
Chat Correction (Beta feature): A feature that corrects things like awkward Korean phrasing, status window output errors, or missing images in the AI's output. It plays a role in further refining and assisting the output of existing models. It's currently offered for free as a beta feature and is being continuously improved.
In the correction guidance, write down the specific direction you want corrected, such as "include at least n images" or "keep the dialect for characters who use one."
If you're additionally using external images not included in the work itself — through persona, user notes, etc. — please also add the rules for that external image in the correction guidance, along with an instruction that that image must always be output as the persona image.
An assistant manager that helps the flow of the story. When activated, it watches the mood of the conversation according to what you want and helps the story develop naturally.Once you activate the Chat Manager, you can simply choose the intervention strength from Defensive, Balanced, or Aggressive. If you want to set things in more detail, you can turn on detailed settings to more directly decide the direction of pacing, mood, and events. If you allow time skips, the AI will quickly move past pointless, dragging blank periods. It's currently offered for free as a beta feature and is being continuously improved.
This is where you make the chat screen more comfortable to view.
Change the font size and font style (default, serif, monospace). If you want a book-like feel, serif is nice.
You can change the chat window width to medium or wide.
You can turn image display on or off, and choose the image size among small, default, or large.
Send with Enter Key (on by default):
Pressing Enter sends the message.
Pressing Shift + Enter creates a new line.(In the chat settings under the viewer menu on the left, turning off the "Send with Enter key" toggle changes it so Enter creates a new line and Ctrl + Enter sends the message.)
Pressing Enter creates a new line.
Messages can only be sent via the send button.
Send Empty Message:When enabled, you can send a message even when the message input field is empty. This is useful when you don't have anything in particular to say, or when you want to see the AI's reply continue just one more time.Separately, pressing the second icon from the left below the AI's reply generates a continuation of the reply without you having to type anything.
The writing style setting changes the style of writing that appears in the chat.
Prompt Version Selection:The default system prompt provided by EdenChat acts as the engine that helps a work's roleplay run smoothly. By default, the prompt the creator selected when making the work is provided as the default.Each prompt version has its own character, but applying a different engine than the default setting chosen by the work's creator may produce output the creator didn't intend.
## V1
- EdenChat's long-verified default prompt. Well-suited to stable responses and a familiar chat flow
- Elevates storytelling quality through literary creative writing and narration
- Specialized for rich, long narration and roleplay with a genre-fiction / web-novel feel
## V1.5-Full
- Keeps V1's creative philosophy while strengthening rules around impersonation prevention and output length
- Focuses on analyzing user input and develops the character's dialogue, actions, and emotional arc scene by scene
## V1.5-Core
- A compressed version that keeps only the minimal core rules needed to build the system
- Lets creators freely set output format and roleplay rules however they want when building a workImpersonation Prevention Prompt: When the AI describes your (the human's) actions, thoughts, or words as if they were its own narration, this is called impersonation. Depending on your preference, you can choose whether to let the AI narrate your activity on your behalf or not. When the impersonation prevention prompt is added, the AI will not narrate your activity.(However, in situations where the default output character count is large but the content of your input message to reference is short, the AI may end up narrating your activity to fill out the content, regardless of whether this feature is enabled.)
Persona
User Notes
Long-term Memory