Voxta Chat
The conversation surface — type or speak, regenerate, inspect, navigate with shortcuts.
The Chat screen is where you actually talk with a character. Either type into the input bar or speak — Voxta listens, transcribes, generates a reply, and speaks it back. Emoji emoticons appear above the avatar based on the conversation tone.
Chat views
Switch views from the chat header, with shift+alt+1 through 5, or from a script with the SelectView app trigger:
| View | What it is | |
|---|---|---|
| Talk | shift+alt+1 | Voice-centric. Minimal UI, mostly just the active speaker. |
| Portrait | shift+alt+2 | Avatar-focused. |
| Chat | shift+alt+3 | Full scrollable history. |
| Assistant | shift+alt+4 | Compact layout for assistant-style use. |
| Debug | shift+alt+5 | Inspector-heavy view for scenario work. |
| Stage | — | The visual-novel surface: background, buttons, HUD. Set by the scenario. |
Chat view is where you get detailed control — message regeneration, editing, and deletion, which the simpler views don't show.
Multi-character chats
When a scenario assigns multiple characters, Chat view + the Quick-access menu give you the cleanest "who said what" view. Each character speaks in turn, the scenario decides who replies next (or your script can force the turn with e.chatFlow()).
Chat Inspector
Open it with shift+alt+i, or Show chat inspector in Chat view. This is Voxta's "what is the AI actually seeing right now?" panel, and the first place to look when a scenario misbehaves.
It is one scrolling panel of sections:
| Section | Shows |
|---|---|
| Audio settings | The live audio configuration for this chat. |
| Flags | Every flag currently set, and when it was set. |
| HUD | The HUD state the stage is rendering. |
| Actions | Candidate actions grouped by layer — what could fire, and what a flag filtered out. |
| Tools | The tools offered to the character this turn. |
| Contexts | The contexts resolved into this prompt. You can add one here temporarily to test a reaction without editing the scenario. |
| Memory | What has been recalled and summarized. |
| Script | The script log, with a code-snippet browser. |
| Chat | Token count and live status — generating, talking, recording. |
| Advanced | Session and chat ids, auto-reply, and the send options below. |
Advanced options
- Enable character replies — off lets you send messages that only run action inference, with no reply.
- Enable character actions — off skips action inference entirely.
- Auto reply — send a set line, or continue the last message, after a delay since the last interaction. This is what lets a story scenario run on its own.
- Show filtered — reveals the actions and contexts currently hidden by their flag conditions.
Both send options affect this client only.
Keyboard shortcuts
| Shortcut | What it does |
|---|---|
shift+alt+m | Toggle the left panel |
shift+alt+i | Open the Chat Inspector |
shift+alt+s | Open the Services tab in the right panel |
shift+alt+h | Toggle visibility of hidden messages (notes, secrets, instructions) |
shift+alt+r | Regenerate the last reply |
shift+alt+t | Speak the last message again |
shift+alt+x | Stop the chat and return home |
shift+alt+k | Skip the audio of the currently-playing message |
shift+alt+1…5 | Switch view |
Use shift+alt+h to see what the AI sees that you normally can't — secrets, notes, instructions, hidden message metadata. Indispensable when debugging scenarios.