Test drive
The debug chat inside the editor, and what it does and does not run.
The character editor has a floating AI Studio button in the bottom-right corner. Its Test tab chats with the draft exactly as the form currently stands — unsaved edits included — so you can hear the character before anyone else does.
It unlocks as soon as the form has a name and a description. Until then the tab is disabled with "Add a name and description before test chat. The creation assistant is still available."
What it runs against
The panel is labelled Uses current form data · not saved. Nothing you type here is stored as a conversation, and nothing about the draft is saved by testing it.
| In the prompt | Not in the prompt |
|---|---|
| Every field currently in the form | Plugins — including ones you pinned as character defaults |
| The scenario you pick, if any | Stickers |
| The persona you pick, if any | |
| When you are editing a saved character: your own memories of it, and the always-on entries of attached world books |
Because plugins never run here, a character built around a tool plugin will behave differently in a real chat. Test that part after saving, in an ordinary conversation.
Limits
| Value | |
|---|---|
| Turns per conversation | 5, shown as a counter above the input |
| Characters per message | 1,000 |
| Model | Fixed. There is no model picker |
At the limit you get "Turn limit reached. Clear to start over." Clear resets the conversation and the counter.
Five turns is enough to catch the things that matter early: does the character answer everything immediately, does it narrate the player's actions, does it sound like the examples you wrote. It is not enough to judge how a long chat drifts.
Controls
- Scenario picker, defaulting to No scenario — the same default a real chat has.
- Persona picker, defaulting to No persona. Pick one of your own personas to see what the character says to that player.
- Per message: Copy, Regenerate, Edit, Delete (which removes that message and everything after it), and arrows to move between regenerated variants.
- Stop interrupts a reply in progress.
The panel deliberately shows no token counts, no prompt preview and no per-message cost.
What it costs
| Who you are | What you pay |
|---|---|
| Subscriber | Nothing, until you have used 30 AI generations in a rolling 5-hour window — an allowance shared with the rest of Reverie's creator AI. The panel then says how long until it resets |
| Everyone else | Credits per reply |
The other two tabs
AI Studio has three tabs, and the other two are worth knowing about while you are in there:
- Create — describe a change in plain language ("make her warmer", "add a rival scenario") and get back a field-by-field diff with before and after values. Apply them one at a time, Apply all, or Undo last AI change. Nothing reaches the form until you apply it.
- Check — the Character experience check: five dimensions (Identity, Distinct voice, Opening experience, Playability, Consistency), each marked Ready, Developing or Not started, each with an Improve with AI action. It is advice about playability, not a requirement to fill every field.
After a test reply you also get feedback chips — Out of character, Too long, Spoke for me, Wrong tone, or your own wording — which turn into a change request on the Create tab.
Elsewhere
The conversational builder has its own test chat for the draft it is holding. The visual builder does not test — it hands you a finished draft to review in the editor.