SpicyChat AI chat controls: messages, saved branches and groups

Starting and continuing a conversation

Selecting a character opens the most recent conversation associated with that profile. A fresh-chat option begins another thread while preserving earlier saved material. Before replying, read the greeting because it establishes the scene and often signals whether the creator expects narration, direct speech or a specific relationship. If the premise is a poor fit, return to the character search guide instead of forcing an unrelated direction.

A useful opening adds information the character can act on: where the scene begins, what the user does and what outcome is being explored. The model still chooses its reply, so concise direction is usually more reliable than a large block of contradictory constraints. Never include real secrets merely to make a fictional profile feel personal.

SpicyChat also documents Group Chats with two to ten chatbots in one shared conversation. Public groups can be discovered from the homepage, and users can create public, unlisted or private groups subject to character visibility. Free users manually select the next responder; premium users may have an automatic responder option. Group Chat currently has separate media limitations, so verify the live group interface before expecting one-to-one chat features.

Actions, formatting and editing

Official documentation describes action notation using text between asterisks, allowing a message to combine narration and speech. Markdown-style formatting can add emphasis or structure. These conventions help readability but do not create a separate secure channel or system instruction; everything remains part of the conversational context interpreted by the selected model.

Editing a generated message is a strong recovery tool. It lets the user remove an unwanted claim, repair tone or redirect the next response. Use it early: later turns may already refer to the problematic text. Message ratings send quality feedback to the platform, but editing is the immediate way to shape the local branch.

Saved chats, clones and Persona changes

Saved conversations make it possible to return to a prior thread. Cloning creates an alternate branch so one path can continue without modifying the original. Give clones descriptive titles that identify the decision point, especially when several branches use the same character. This is more reliable than assuming the chat list will explain why two similar transcripts diverged.

A Persona changes the identity and highlights presented for future interaction. Official guidance notes that switching Persona mid-conversation does not rewrite references already stored in earlier messages or memories. Plan important identity changes near a clean branch, and review the Persona and memory guide before expecting a new profile to erase established context.

Response length, images and model limits

Response length and context depend on the model and plan. A longer answer can provide richer prose while also increasing the chance of drift, repetition or unrequested detail. More context gives the model access to a larger conversation window; it does not transform generated statements into facts. Correct important details in the transcript and keep a short memory summary when a scene becomes complex.

Conversation images and advanced model choices are described as plan-dependent features. Generated images can also misinterpret a scene or character description. Check live availability in the premium features guide and avoid buying a tier solely because one example output looked strong.