Node Flux Text-to-Speech

Get started using Deepgram's Flux streaming text-to-speech (v2 speak) with this Node demo app.

Unlike the Flux transcription starter (node-flux), which is a raw WebSocket proxy, this starter uses the official @deepgram/sdk speak.v2 client on the backend — so the SDK manages the Deepgram WebSocket, auth, and binary-audio framing for you.

How it works

Browser  ──(JSON control: Speak / Flush / Close)──▶  Node backend  ──(@deepgram/sdk speak.v2)──▶  Deepgram Flux TTS
Browser  ◀──(binary audio frames + JSON control)───  Node backend  ◀──(@deepgram/sdk speak.v2)───  Deepgram Flux TTS

• The backend never exposes your API key — the browser authenticates to the backend with a short-lived JWT (/api/session), passed on the WebSocket access_token.<jwt> subprotocol.
• The backend opens one client.speak.v2 connection per browser client and bridges messages both ways.
• Audio frames from Deepgram arrive as binary and are forwarded to the browser as binary; control messages (Connected, SpeechStarted, SpeechMetadata, Flushed, Warning, Error, …) are forwarded as JSON.

Client → backend message protocol

Send JSON text frames on the /api/tts WebSocket:

{ "type": "Speak", "text": "Hello from Flux." }   // synthesize text
{ "type": "Flush" }                                // finish the current turn, flush audio
{ "type": "Close" }                                // end the session

Prerequisites

• Node.js 24+ and pnpm
• A Deepgram API key — get one at console.deepgram.com.

Local Development

Makefile (Recommended)

make init
cp sample.env .env   # add your DEEPGRAM_API_KEY
make start           # backend on http://localhost:8081

Node.js & pnpm

git clone https://github.com/deepgram-starters/node-flux-tts.git
cd node-flux-tts
corepack pnpm install
cp sample.env .env   # add your DEEPGRAM_API_KEY
node --no-deprecation server.js

Configuration

Environment variables (see sample.env):

Variable	Default	Description
DEEPGRAM_API_KEY	—	Required. Your Deepgram API key.
PORT	8081	Backend port.
HOST	0.0.0.0	Backend host.
DEEPGRAM_TTS_MODEL	flux-alexis-en	Flux voice (flux-{voice}-{language}).
SESSION_SECRET	random per boot	Set in production for stable JWT signing.

The connection also accepts model, encoding, and sample_rate as query params on the /api/tts WebSocket.

License

MIT - See LICENSE