Voice Chat SDKs

Add AI voice chat to your game or app. Record, send, hear back - one API call.

Download

Unity

C# MonoBehaviour

Drop-in component with Microphone recording, WAV encoding, UnityWebRequest upload, and AudioClip playback.

Download .cs

Godot

GDScript Node

AutoLoad-ready script with AudioEffectRecord capture, manual multipart upload, and AudioStreamWAV playback.

Download .gd

Unreal Engine

C++ Actor Component

Blueprint-friendly component with AudioCapture, WAV encoding, FHttpModule multipart upload, and SoundWave playback.

.h .cpp

Next.js / React

TypeScript Hook + Component

useSparkVoiceChat hook + drop-in <SparkVoiceChat> component with Web Audio API recording and streaming playback.

Download .tsx

How It Works

Every SDK follows the same flow. One HTTP call, three AI steps happen server-side:

Record audio

→

POST to API

→

Play response

On the server, your audio goes through three AI-powered steps:

Whisper STT

→

Chat AI

→

Orpheus TTS

All three steps are handled by SparkAI. No additional keys or services needed.

Getting Started

Before using any SDK, you need an API key:

Create an account

Create an app

In the dashboard, click "Create App" and choose API / Extension as the app type. Pick your platform (Chrome Extension, CLI Tool, or npm Package).

Copy your API key

Your key starts with spark_ai- and is shown only once. Copy it immediately.

Where to store your key

Never hardcode your API key in source code. Each platform has a secure storage method:

Platform	Recommended Storage
Next.js / React	Add `SPARK_AI_KEY=spark_ai-...` to `.env.local` and read via `process.env.SPARK_AI_KEY` in a server-side API route. Never expose in client-side code.
Node.js / CLI	Export as `SPARK_AI_KEY` in your shell profile or `.env` file. Read via `process.env.SPARK_AI_KEY`.
npm Package	Accept the key as a constructor parameter. Document that consumers should set `SPARK_AI_KEY` in their `.env`.
Chrome Extension	Store in `chrome.storage.local` or inject at build time via your bundler. Add `.env` to `.gitignore`.
Unity	Store in a `ScriptableObject` asset excluded from version control, or load from a config file at runtime.
Godot	Use an `@export` variable set in the Inspector, or load from a `.cfg` file excluded from version control.
Unreal Engine	Set via the Details panel at the instance level, or load from a config file. Avoid committing keys in source files.

Your API key authenticates via the X-Api-Key header. All rate limits and usage are tracked per key.

API Reference

POST /api/voice-chat

Send audio, receive AI audio response. Requires a paid plan.

Request Headers

Header	Value
`X-Api-Key`	Your API key (`spark_ai-...`)	REQUIRED
`Content-Type`	`multipart/form-data`	REQUIRED

Request Body (Form Data)

Field	Type	Description
`audio`	File	WAV audio file, max 10MB REQUIRED
`sessionId`	String	Conversation session ID (for multi-turn context)
`userEmail`	String	User identifier (for per-user rate limits)

Response (200 OK)

Header	Description
`Content-Type`	`audio/wav` - raw WAV audio bytes in body
`X-Transcript`	URL-encoded transcript of what the user said
`X-AI-Response`	URL-encoded text of the AI response

cURL Example

Shell

curl -X POST https://sparkbrain.app/api/voice-chat \
  -H "X-Api-Key: $SPARK_AI_KEY" \
  -F "[email protected]" \
  -F "sessionId=my-session" \
  --output response.wav

Unity

Add SparkVoiceChat.cs to any GameObject. Set your API key in the Inspector or load from a ScriptableObject - never hardcode it.

C# - Setup

var voice = gameObject.AddComponent<SparkVoiceChat>();
voice.apiKey = myConfig.sparkAiKey; // load from ScriptableObject
voice.endpoint = "https://sparkbrain.app/api/voice-chat";

voice.OnTranscriptReceived += (t) => Debug.Log("You said: " + t);
voice.OnResponseReceived += (r) => Debug.Log("AI said: " + r);
voice.OnError += (e) => Debug.LogError(e);

C# - Usage

// Start recording (call on button press)
voice.StartRecording();

// Stop and send to AI (call on button release)
voice.StopRecordingAndSend();
// Audio response plays automatically

Events: OnTranscriptReceived, OnResponseReceived, OnAudioPlayed, OnError

Godot

Add spark_voice_chat.gd as an AutoLoad or attach to a Node. Set your API key via the Inspector @export or load from a config file.

GDScript - Setup

# Project Settings > Audio > Driver > Enable Input = true
# Add audio bus "Record" with AudioEffectRecord

var voice = $SparkVoiceChat
voice.api_key = my_config.spark_ai_key  # set in Inspector or load from .cfg
voice.endpoint = "https://sparkbrain.app/api/voice-chat"

voice.transcript_received.connect(func(t): print("You: ", t))
voice.response_received.connect(func(r): print("AI: ", r))
voice.error.connect(func(e): push_error(e))

GDScript - Usage

# Start recording
voice.start_recording()

# Stop and send to AI
voice.stop_and_send()
# Audio plays automatically via AudioStreamPlayer child

Signals: transcript_received, response_received, audio_played, error

Unreal Engine

Add the SparkVoiceChat component to any Actor. Enable the AudioCapture plugin.

C++ - Setup

// In your Actor's header
UPROPERTY(VisibleAnywhere)
USparkVoiceChat* VoiceChat;

// In BeginPlay() - set ApiKey via Details panel or load from config
VoiceChat = NewObject<USparkVoiceChat>(this);
VoiceChat->Endpoint = TEXT("https://sparkbrain.app/api/voice-chat");
VoiceChat->RegisterComponent();

VoiceChat->OnTranscriptReceived.AddDynamic(this, &AMyActor::HandleTranscript);
VoiceChat->OnResponseReceived.AddDynamic(this, &AMyActor::HandleResponse);

C++ - Usage

// Start recording
VoiceChat->StartRecording();

// Stop and send to AI
VoiceChat->StopRecordingAndSend();
// Audio plays automatically via PlaySound2D

All properties are Blueprint-exposed. Set ApiKey in the Details panel instead of hardcoding in source.

Next.js / React

Use the useSparkVoiceChat hook or the drop-in <SparkVoiceChat> component. Add SPARK_AI_KEY=spark_ai-... to your .env.local and pass it from a server component or API route.

TypeScript - Hook

import { useSparkVoiceChat } from './SparkVoiceChat';

function VoiceChat({ apiKey }: { apiKey: string }) {
  const { startRecording, stopAndSend, isRecording, isProcessing,
          transcript, aiResponse, error } = useSparkVoiceChat({
    apiKey,  // pass from server, never hardcode
    endpoint: 'https://sparkbrain.app/api/voice-chat',
  });

  return (
    <button onClick={isRecording ? stopAndSend : startRecording}
            disabled={isProcessing}>
      {isProcessing ? 'Thinking...' : isRecording ? 'Stop' : 'Speak'}
    </button>
  );
}

TypeScript - Component

import { SparkVoiceChat } from './SparkVoiceChat';

// Drop-in with built-in UI
<SparkVoiceChat
  apiKey={process.env.SPARK_AI_KEY!}
  endpoint="https://sparkbrain.app/api/voice-chat"
/>

The hook gives full control. The component provides a ready-made button with transcript display. Keep your key in .env.local, not in source code.