Twilio Voice + Gemini ConversationRelay Hub

🔗 Twilio Webhook URL (TwiML):

https://.../twiml

⚡ ConversationRelay WebSocket:

wss://.../ws

Live Call Stream

Real-time transcripts, token streaming, and caller interruptions

No Active Calls

📞

Waiting for Calls

Dial into your Twilio phone number, or launch the In-Browser Simulator to start chatting. Spoken words and Gemini streamed responses will appear here in real-time.

Call History

Recent sessions & metrics

No calls recorded yet.

Twilio Relay • Gemini AI

00:00

🤖

Gemini Voice Assistant

Model: gemini-2.5-flash

Ready to Call

Press "Start Call" or type a message below to test your assistant.

Simulator Model Switcher

Switch models on the fly during this simulated call:

Gemini 3.1 Flash Live Preview Live API

Sub-200ms ultra low-latency real-time bidirectional stream.

Gemini 2.5 Flash Voice

Conversational chat model with fast streaming response.

Gemini 3.6 Flash Latest

Google's latest recommended high-speed reasoning model.

Gemini 2.5 Pro Pro

Maximum intelligence for tricky queries.

Play voice aloud in browser (Auto-detects Hebrew & English) Auto-listen via Microphone when bot finishes speaking

Speech Recognition Language:

Packet Inspector

Emulated Twilio ConversationRelay WebSocket packets:

[System] Simulator initialized. Ready to connect.

Gemini 3.5 Transcribe

Speech-to-text powered by gemini-3.5-transcribe with Hebrew understanding and 85+ languages.

Gemini 3.5 ASR

Language Hint:

Transcription Model:

00:00

Click to speak in Hebrew or English

📁

Drag & drop audio file here, or Browse

Supported: WAV, MP3, M4A, OGG, WebM (up to 25MB)

Transcription Output

Recognized speech output with Hebrew & English formatting:

Awaiting Audio

📝

No Transcription Yet

Record your voice or upload an audio file and click "Transcribe with Gemini 3.5" to see the transcribed text.

Gemini Model Configuration

Select the default Gemini model used for incoming Twilio phone calls.

Or specify a custom Gemini model ID:

Google Gemini API Key

Your key is stored securely in .env on your machine.

Gemini API Key:

Voice Persona & System Prompt

Define how the assistant introduces itself and speaks to callers.

Prompt Preset:

System Instruction: Optimized for phone TTS: rules to spell numbers and avoid emojis/markdown are automatically enforced.

Welcome Greeting (Spoken on Answer):

Text-to-Speech (TTS) & Language Settings

Twilio ConversationRelay voice synthesis and recognition parameters:

TTS Provider:

Voice Recognition Language (STT):

Suggested Voices:

Active Voice ID:

Temperature (0.0 - 1.0):

Public Host / Domain Override (Optional):

Twilio Account Credentials

Enter your Twilio Account SID and Auth Token to connect Twilio Voice ConversationRelay.

Not Connected

Twilio Account SID *: Found on the main dashboard of your Twilio Console.

Twilio Auth Token *:

Secret authentication token associated with your Account SID.

Phone Number & Webhook Setup

Select your Twilio phone number and link it to this server with 1 click.

Select Active Twilio Phone Number:

Or enter manually:

Public Server Webhook URL:

https://34.41.149.133.sslip.io/twiml

Method: HTTP POST • WebSocket: wss://34.41.149.133.sslip.io/ws

Manual Setup Instructions (Twilio Console)

Go to Phone Numbers > Manage > Active Numbers in the Twilio Console.
Click on your phone number.
Under Voice Configuration, set A CALL COMES IN to Webhook.
Paste the URL: https://34.41.149.133.sslip.io/twiml with HTTP POST.
Click Save.

TwiML XML Preview

ConversationRelay instruction returned to Twilio Voice:

<!-- Loading TwiML... -->

📞

Option A - Inbound Call: Dial +12082853907 from your mobile phone (ensure you dial with +1 or 001 for international calling). Gemini will answer and speak with you!

Option B - Twilio Calls You:

Enter your phone number with country code, and Twilio will ring your phone directly: