Decide what you want from the first exchange
Give your character an immediate reason to speak. After checking the description and greeting, establish where you are and what you are doing there. You need enough detail for the other character to respond, not an outline that settles the whole story in advance.
Imagine bringing an unopened parcel to a village post office because its address names a street that no longer exists. You can explain where you found it and ask the clerk for help. The question begins a scene without deciding what the clerk knows or what is inside.
Choose the kind of change the reply needs
Use Continue when you want the response to carry on from its current point. For a wrong name or an action you did not intend, editing may be enough. Regeneration gives you another version when the response as a whole does not fit. Extending a response is different from sending a new instruction.
If the conversation keeps describing the same moment, give it a development to react to. A delivery arrives, an arrangement falls through or someone asks an urgent question. Even a small change can give the next exchange a purpose beyond repeating the surroundings.
Keep different versions of the story distinct
Before changing an older story, locate the conversation you mean to use. Chats takes you back to recent threads, while a character's saved-chat list can contain other conversations with that same character. The latest thread is not necessarily the one you were looking for.
Choose a new conversation for a new starting premise. If you want to keep the current story but test a different outcome, cloning creates another branch to follow. Confirm which version is open before removing messages or making substantial edits.
Questions
Can I keep different storylines with one character?
Yes. Separate saved chats can hold different premises and outcomes for the same character. To find an earlier version, check the saved-chat list rather than relying only on the most recent conversation.
Do people write the messages I receive from a character?
No. We generate character dialogue using language models. An in-character claim to be human does not mean that a real person has joined your conversation.