Voice Nodes Reference

Voice nodes enable phone-based automation, from handling inbound calls to making AI-powered outbound calls and voice responses.

---

Overview

Voice workflows typically follow this pattern:

[Voice Trigger] → [Voice Agent] → [Actions] → [Voice Response/Transfer]

Node Categories

Category	Nodes
Triggers	Voice Inbound, Voice Message
AI Agent	Voice Agent
Actions	Voice Outbound, Transfer Call
Output	Voice Response, Text to Speech

Supported Providers

• Telephony: Exotel, Twilio, Tata Tele
• TTS (Text-to-Speech): ElevenLabs, Cartesia, Azure, Google, AWS Polly
• STT (Speech-to-Text): Deepgram, Azure, Google, AWS Transcribe

---

Voice Inbound Trigger

Triggers workflows when incoming phone calls are received.

Configuration

Field	Type	Required	Description
Name	Text	No	Name for this trigger
Voice Provider	Select	Yes	Phone provider integration (Twilio, Exotel)
Phone Numbers	Multi-select	Yes	Phone numbers that trigger this workflow
Enable Recording	Toggle	No	Enable call recording
Call Timeout	Number	No	Max call duration (30-3600 seconds)
Business Hours Only	Toggle	No	Only trigger during business hours

Business Hours Configuration

When Business Hours Only is enabled:

Timezone: UTC
Monday:    09:00 - 17:00 (enabled)
Tuesday:   09:00 - 17:00 (enabled)
Wednesday: 09:00 - 17:00 (enabled)
Thursday:  09:00 - 17:00 (enabled)
Friday:    09:00 - 17:00 (enabled)
Saturday:  10:00 - 14:00 (disabled)
Sunday:    (disabled)

Output Variables

{{voiceinboundtrigger_1.caller_number}}    - Caller's phone number
{{voiceinboundtrigger_1.called_number}}    - Number that was called
{{voiceinboundtrigger_1.call_id}}          - Unique call identifier
{{voiceinboundtrigger_1.call_start_time}}  - When call started
{{voiceinboundtrigger_1.provider}}         - Voice provider name

---

Voice Message Trigger

Triggers workflows on voice input from various sources (phone, WebRTC, WhatsApp, voice notes).

Configuration

Field	Type	Required	Description
Source	Select	No	Filter by source: any, phone, webrtc, whatsapp, voice_note
Test Mode	Toggle	No	Allow text input for testing
Transcriber Source	Select	No	direct, voice_agent, or none
Transcriber Integration	Select	Conditional	STT provider (when source is direct)
Language	Text	No	Language code (default: en)
Include Timestamps	Toggle	No	Include word-level timestamps
Filter Profanity	Toggle	No	Filter profanity from transcription

Output Variables

{{voicemessagetrigger_1.message}}          - Transcribed text
{{voicemessagetrigger_1.audio_url}}        - Original audio URL
{{voicemessagetrigger_1.source}}           - Input source
{{voicemessagetrigger_1.confidence}}       - Transcription confidence
{{voicemessagetrigger_1.duration}}         - Audio duration
{{voicemessagetrigger_1.language}}         - Detected language

---

Voice Agent Node

Voice-enabled AI agent for handling phone conversations. Similar to Core Agent but optimized for voice interactions.

Configuration

Basic Settings

Field	Type	Required	Description
Name	Text	No	Display name for the agent
LLM Provider	Select	Yes	AI provider (OpenAI, Anthropic, etc.)
Temperature	Slider	No	Creativity (0-1, default 0.7)
Max Tokens	Number	No	Response length (50-500, default 250)
System Prompt	Textarea	Yes	Agent instructions

Voice Settings (TTS)

Field	Type	Required	Description
Synthesizer Integration	Select	Yes	TTS provider
Voice ID	Select	Yes	Voice to use
Speed	Slider	No	Speech speed (0.5-2.0)
Language	Text	No	Voice language
Accent	Text	No	Voice accent

Provider-Specific Settings:

ElevenLabs:

• Stability (0-1)
• Similarity Boost (0-1)
• Style (0-1)

Cartesia:

• Model variant

Transcriber Settings (STT)

Field	Type	Required	Description
Transcriber Integration	Select	Yes	STT provider
Language	Text	No	Transcription language
Model	Select	No	STT model (Deepgram)
Profanity Filter	Toggle	No	Filter profanity
Smart Formatting	Toggle	No	Format numbers, dates
Speaker Diarization	Toggle	No	Identify speakers

Timing Settings:

• Silence Timeout: 100-5000ms (when to finalize transcript)
• Max Segment Duration: 20-70s
• Utterance Timeout: 500-3000ms

Channel Configuration

Field	Type	Description
Channel	Select	phone, webrtc, or auto
Voice Provider	Select	Telephony provider for phone calls

Output Variables

{{voiceagent_1.output}}              - Final agent response
{{voiceagent_1.transcript}}          - Full conversation transcript
{{voiceagent_1.call_duration}}       - Call duration
{{voiceagent_1.success}}             - true/false
{{voiceagent_1.tool_calls}}          - Tools used during call

---

Voice Outbound Call Node

Make outbound phone calls - either manual (connect two humans) or AI-powered.

Configuration

Call Type

Type	Description
Manual	Connect two humans (e.g., agent to customer)
AI	AI Voice Agent handles the call

Core Settings

Field	Type	Required	Description
Call Type	Select	Yes	manual or ai
Voice Provider	Select	Yes	Telephony integration
Caller ID	Text	Yes	Display number for customer
Call Input Mode	Select	No	single or batch

Single Call Input

Phone: {{contact.phone}}
Variables: [
  {key: "name", value: "{{contact.first_name}}"},
  {key: "company", value: "{{contact.company.name}}"}
]

Batch Call Input

Data Source: {{databasequery_1.contacts}}
Phone Field: phone
Variable Mappings: [
  {key: "name", value: "first_name"},
  {key: "deal_value", value: "deal.value"}
]

Manual Call Configuration

Field	Type	Description
From Number	Text	Agent's phone number
Connect Order	Select	from_first (call agent first) or to_first (call customer first)
Call Purpose	Text	Notes for CRM logging

AI Call Configuration

Field	Type	Description
Voice Agent Node	Select	Voice Agent to handle call
Conversation Variables	Key-Value	Variables to pass to agent

Output Variables

{{voiceoutbound_1.call_id}}          - Unique call ID
{{voiceoutbound_1.status}}           - Call status
{{voiceoutbound_1.duration}}         - Call duration
{{voiceoutbound_1.recording_url}}    - Recording URL (if enabled)
{{voiceoutbound_1.success}}          - true/false

# For batch calls
{{voiceoutbound_1.total_calls}}      - Total calls attempted
{{voiceoutbound_1.successful}}       - Successful calls
{{voiceoutbound_1.failed}}           - Failed calls
{{voiceoutbound_1.results}}          - Array of call results

---

Transfer Call Node

Transfer the current call to another phone number or department. Works as a tool for Voice Agent.

Configuration

Field	Type	Required	Description
Destinations	Array	Yes	List of transfer destinations
Transfer Message	Text	No	Message before transfer
Fallback Message	Text	No	Message if transfer fails

Destination Configuration

[
  {
    "name": "Sales Team",
    "number": "+919876543210",
    "description": "For pricing and purchase inquiries"
  },
  {
    "name": "Support",
    "number": "+919876543211",
    "description": "For technical issues and troubleshooting"
  },
  {
    "name": "Billing",
    "number": "+919876543212",
    "description": "For payment and invoice questions"
  }
]

How It Works

1. Voice Agent recognizes transfer intent from conversation
2. AI calls this tool with the destination name
3. Node speaks transfer message and initiates transfer
4. If transfer fails, fallback message is played

Supported Providers

• Exotel: Passthru → Connect flow
• Tata Tele: API-based transfer
• Twilio: SIP REFER or REST API

Output Variables

{{transfercall_1.transferred}}       - true/false
{{transfercall_1.destination}}       - Destination transferred to
{{transfercall_1.transfer_time}}     - When transfer occurred

---

Voice Response Node

Convert text to speech and send as audio response (primarily for active calls).

Configuration

Field	Type	Required	Description
Text	Text	Yes	Text to speak (supports variables)
Synthesizer Source	Select	Yes	direct, voice_agent, or none
Synthesizer Integration	Select	Conditional	TTS provider
Output Format	Select	No	mp3, wav, ogg, m4a
Output Destination	Select	No	file, stream, phone, webrtc

Voice Settings

Setting	Range	Default	Description
Voice ID	-	-	Voice to use
Speed	0.5-2.0	1.0	Speech speed
Pitch	-20 to 20	0	Voice pitch
Volume	0.0-1.0	0.8	Audio volume

Output Variables

{{voiceresponse_1.audio_url}}        - URL to audio file
{{voiceresponse_1.audio_data}}       - Audio binary (base64)
{{voiceresponse_1.duration}}         - Audio duration
{{voiceresponse_1.success}}          - true/false

---

Text to Speech Node

Convert text to speech audio file (for generating audio assets, not live calls).

Configuration

Field	Type	Required	Description
Synthesizer Integration	Select	No	TTS provider
Voice ID	Text	No	Voice to use
Language	Text	No	Language code (e.g., en-US)
Voice Gender	Select	No	male, female, neutral, auto
Speaking Rate	Slider	No	Speed (0.5-2.0)
Pitch	Slider	No	Pitch (-20 to 20)
Volume	Slider	No	Volume (0-100)
Audio Format	Select	No	mp3, wav, ogg, opus, flac, aac

Text Source Options

Source	Description
input	From previous node output
static	Fixed text in configuration
template	Template with variables
previous_node	Specific node's output

Advanced Options

Field	Description
Use SSML	Enable SSML markup for advanced control
Output Mode	base64, url, file, or stream

Output Variables

{{texttospeech_1.audio_data}}        - Audio as base64
{{texttospeech_1.audio_url}}         - URL to audio file
{{texttospeech_1.duration}}          - Audio duration
{{texttospeech_1.format}}            - Audio format used
{{texttospeech_1.success}}           - true/false

---

Common Workflow Patterns

Inbound Call → AI Agent

[Voice Inbound Trigger]
    ↓
[Voice Agent]
  - System: "You are a helpful customer service agent..."
  - Tools: [KB Search, CRM Manager, Transfer Call]
    ↓
[CRM Manager: Log Activity]

AI Outbound Campaign

[Database Query: Get contacts]
    ↓
[Voice Outbound Call]
  - Call Type: AI
  - Batch Mode
  - Voice Agent: Sales Agent
    ↓
[CRM Manager: Update status]

Manual Call Connect

[Click Trigger: "Call Customer"]
    ↓
[CRM Manager: Get Contact]
    ↓
[Voice Outbound Call]
  - Call Type: Manual
  - From: {{user.phone}}
  - To: {{contact.phone}}

Voice Note Processing

[Voice Message Trigger]
  - Source: voice_note
    ↓
[LLM Integration: Summarize]
    ↓
[CRM Manager: Create Note]
    ↓
[Text Response: "Note saved!"]

IVR with Transfer

[Voice Inbound Trigger]
    ↓
[Voice Agent]
  - "How can I help today?"
  - Tools: [Transfer Call, KB Search]
    ↓
[Condition: Transfer requested?]
    → Yes → [Transfer Call]
    → No → [Voice Response: "Goodbye!"]

---

Best Practices

Voice Agent Prompts

Keep prompts concise for voice:

You are a friendly customer service agent for Acme Corp.
- Keep responses under 2-3 sentences
- Speak naturally, avoid technical jargon
- Confirm understanding before actions
- Offer to transfer to human if needed

Transcription Settings

For real-time conversations:

• Silence Timeout: 300ms (faster response)
• Max Segment: 30s
• VAD Threshold: 0.03 (balanced)

Call Recording

Always inform callers about recording:

Transfer Message: "This call may be recorded for quality purposes.
Please hold while I transfer your call."

Error Handling

[Voice Agent] → [Condition: Success?]
             → Yes → [Log: Call completed]
             → No → [Voice Response: Apology] → [Create Task: Follow up]

Batch Call Limits

• Start with small batches (10-20 calls)
• Monitor success rates before scaling
• Implement retry logic for failed calls
• Respect carrier rate limits