Skip to content

AI Voice Chat Demo

Two sample scripts demonstrating the AI voice and text chat system — persona-based conversations with real-time captions, microphone recording, and standalone TTS/STT services.

Scene Overview

Location: Assets/_IntelliVerseXSDK/Demos/IVXAIVoiceChatDemo.cs and IVXAIVoiceServicesDemo.cs

This sample demonstrates:

  • AI persona selection (Fortune Teller, Teacher, Coach, etc.)
  • Timed voice/text chat sessions with real-time captions
  • Microphone recording, commit, and playback
  • Text-to-Speech synthesis
  • Speech-to-Text streaming transcription
  • Language detection from audio
  • Available voice listing

Scene Hierarchy — Voice Chat

Canvas
├── Background
├── Root
│   ├── PersonaGrid
│   │   ├── Title ("Choose Your AI Persona")
│   │   └── GridHolder (2-column card grid)
│   │       ├── Fortune Teller
│   │       ├── AI Teacher
│   │       ├── Career Coach
│   │       ├── Story Teller
│   │       ├── Party Host
│   │       └── Health Advisor
│   └── ChatPanel
│       ├── Header (back, title, timer, end)
│       ├── ScrollArea (chat bubbles)
│       ├── CaptionBar (live captions)
│       └── InputBar (mic, text input, send)

Key Components

IVXAIVoiceChatDemo — Persona Chat

// Starting a voice session with a persona
IVXAISessionManager.Instance.OnCaptionReceived += (text) =>
    captionLabel.text = text;
IVXAISessionManager.Instance.OnCaptionComplete += (text) =>
    AddBubble(text, isUser: false);

IVXAISessionManager.Instance.StartVoiceSessionDirect("Fortune Teller",
    onSuccess: resp =>
    {
        sessionDuration = resp.DurationSeconds;
        AddBubble("Hello! How can I help?", isUser: false);
    },
    onError: err => AddBubble($"[Error: {err}]", isUser: false));

Sending Text Messages

IVXAISessionManager.Instance.SendText("What does my future hold?");

Microphone Recording

// Start recording
IVXAISessionManager.Instance.StartRecording();

// Stop and send audio
IVXAISessionManager.Instance.StopRecording();
IVXAISessionManager.Instance.CommitAudio();

Ending a Session

IVXAISessionManager.Instance.EndVoiceSession();

IVXAIVoiceServicesDemo — Standalone Services

var svc = IVXAIVoiceServices.Instance;

// Text-to-Speech
svc.SynthesizeSpeech("Hello world", voiceId: null, audioBytes =>
    Debug.Log($"Synthesized {audioBytes.Length} bytes"));

// List available voices
svc.ListVoices(voices =>
{
    foreach (var v in voices) Debug.Log(v);
});

// Streaming Speech-to-Text
svc.StartStreamingTranscription();
// ... user speaks ...
svc.StopStreamingTranscription();
svc.TranscribeAudio(audioData, sampleRate: 16000, result =>
    Debug.Log($"Transcription: {result}"));

// Language detection
svc.DetectLanguage(audioData, sampleRate: 16000, (lang, confidence) =>
    Debug.Log($"Detected: {lang} ({confidence:P0})"));

How to Use

Running the Voice Chat Demo

  1. Add IVXAIVoiceChatDemo to any Canvas
  2. Optionally configure IVXAISessionManager in the scene for live backend
  3. Press Play — the persona selection grid appears
  4. Tap a persona card to start a chat session

Testing Without a Backend

The demo operates in mock mode when IVXAISessionManager is not initialized — messages receive placeholder responses so you can test the UI flow.

Testing Voice Services

  1. Add IVXAIVoiceServicesDemo to any GameObject
  2. Ensure IVXAIVoiceServices is initialized with an IVXAIConfig
  3. Use the Speak button for TTS, Start Streaming for STT

Testing the Session Timer

Each session runs on a configurable countdown (default 180 seconds). When the timer reaches zero, EndVoiceSession() is called automatically.

Customization

Adding Personas

Edit the MOCK_PERSONAS and MOCK_ICONS arrays in IVXAIVoiceChatDemo.cs to add new persona cards.

Changing Session Duration

Set _sessionDuration before calling StartVoiceSessionDirect, or modify the backend response's DurationSeconds.

See Also