A companion may know a reader’s preferences in a one-to-one conversation. A group scene introduces another question: which information belongs to every participant?
SoulChat’s product presents groups, while SweetChat’s pricing page describes conversations with two or three characters. SoulChat’s current publicly linked group-creation form permits two to four selected companions, with Start disabled below two. This is a reading of the served interface, not a completed group session or entitlement test.
These counts refer to AI characters, not the human reader. A scene with two companions and you is different from one requiring four companions. If four AI roles are essential, SoulChat has specific public interface evidence; SweetChat’s advertised range does not establish that capacity.
Start with what the group is doing
Consider three fictional adults planning a community exhibition: an artist, an organizer and the reader helping choose a layout. The shared problem is that the display space is smaller than expected.
The group brief can be short:
| Shared information | Purpose |
|---|---|
| The exhibition opens next week | Establishes the same deadline |
| Two display walls are available | Gives everyone the same physical constraint |
| The artist proposes a layout; the organizer explains logistics | Distinguishes the speakers’ responsibilities |
| The reader decides which proposal to discuss | Preserves the reader’s role |
These instructions do not need the reader’s real workplace, address or relationship history. A plausible scene can stay specific while using entirely fictional details.
What stays with an individual character
The artist’s preference for minimal displays is a character trait. The organizer’s concern about visitor flow is another. Neither should be copied into every speaker’s personality.
A general scene instruction such as “name the speaker before each turn” belongs to the group presentation. It should not be written as if the artist alone controls the entire cast.
If all participants sound identical, inspect whether the brief assigned distinct motives. More background will not help if every character has been told to be equally agreeable about every option.
Product evidence and account limits
SweetChat explicitly describes a small cast, but its plan cards and concluding copy disagree about some premium entitlements. We do not infer a universal lowest eligible tier from the presence of a group feature.
SoulChat’s inspected creation surface is a companion selector, with selected cards and a Start count. We did not find a shared-backstory input in that form. Prepare the common premise as the first user message rather than claiming the product has a named group-brief field:
This is a fictional community exhibition. We have two walls and open next week. The artist proposes a layout; the organizer checks visitor flow. Discuss one proposal each, then ask me which to develop. The layout is not decided yet.
That message gives every participant the same explicit task. It is new conversational context, not an import of earlier personal memories. SoulChat’s privacy policy does not establish that a given one-to-one memory automatically becomes available to every group character. The same memory-transfer conclusion cannot be inferred from SweetChat’s cast-size description.
Observe one decision before expanding the cast
A proposed first session can resolve only the display layout. Notice whether speakers remain identifiable, respond to the same constraint and leave the reader’s decision open.
If a reply assigns the reader’s choice or gives one character information that should be private to another, note that specific result. We did not run this test on either service.
Choose around the group surface and current account offer you can actually use. A separate shared brief makes a group understandable; the feature label alone cannot guarantee distinct voices, memory isolation or continuity across future sessions.