Voice Bot AI

Comprehensive Voice AI platform — multi-channel connectivity with Zalo, WhatsApp, Telephone, integrated Inbound & Outbound call center, deployable on Web and Mobile with dedicated SDKs

98%

Recognition Accuracy

<100ms

Real-time Processing Latency

10+

Integrated Channels

24/7

Continuous Operation

Multi-channel Connectivity

A single Voice Bot platform — receives and processes calls from all popular communication channels, syncing data in real-time

Communication Channels

Zalo

Telephone

Viber

Messenger

Web Widget

Backend Systems

CRM

Customer Mgmt

Call Center

Agents

Analytics

Analysis

Ticketing

Requests

Knowledge

API

Integration

Multi-channel Reception

Customers call from any channel — Zalo, WhatsApp, Telephone, Viber, Messenger, or Web Widget

AI Smart Processing

Voice Bot recognizes speech, understands intent, queries data, and responds with natural voice

System Sync

All call data is synced to CRM, Call Center, Analytics, and backend systems

Inbound & Outbound Call Center

Complete AI call center solution for both incoming and outgoing calls, automating 80% of customer service processes

Inbound Voice Bot

Receive & process incoming calls

Voice Bot receives incoming calls from customers, automatically identifies intent, looks up information, and responds with natural voice. When needed, the bot transfers to the appropriate agent with full conversation context.

Incoming Call Processing Flow

Customer calls in

Via SIP Trunk / Zalo / WhatsApp

Bot greets & identifies

STT converts voice → text

NLU analyzes intent

Identifies intent & entities

Process request

Query API, database, knowledge base

Respond with voice

TTS generates natural response

Transfer (if needed)

Escalation to agent + context

Smart IVR

AI-powered automatic call routing system, understanding customer intent through voice instead of key presses

Auto FAQ Response

Bot automatically handles 80% of common questions like order tracking, business hours, return policies

Smart Transfer

When beyond capability, bot transfers calls to the right agent with a conversation summary

Multilingual

Supports Vietnamese, English, Japanese, Korean with native pronunciation for each region

Customer Authentication

Identity verification via voice biometrics or automatic OTP, ensuring security

24/7 Operation

Never miss a call, receive and process anytime including after business hours

80%

Calls auto-processed

No staff needed

<3s

Response time

Average per sentence

95%

Satisfaction rate

CSAT score

60%

Operating cost reduction

vs. traditional call center

No-Code Flow Builder

Two script creation modes — drag-and-drop visual Workflow or natural language description with Prompt Agent. No coding required

flow-builder.antcoai.ai

Auto-saved

Incoming call

STT

Voice → Text

Intent Router

Routing

API Action

Order lookup

Knowledge

Answer FAQ

Escalation

Transfer to agent

TTS Response

Voice response

End

Continue

Drag-and-drop nodes onto canvas

TriggerSTT NodeIntent RouterAPI ActionConditionKnowledge BaseTTS ResponseEscalation

Prompt Agent

Describe scenarios in natural language — AI automatically creates a complete flow. Or let the Agent autonomously handle calls end-to-end

system_prompt.txt

You are a customer service agent of ABC store. When customers ask about orders, call API /orders to look up. If customers want returns, create a ticket and escalate to support. Always respond with a friendly, professional tone.

Natural Language Setup

Describe scenarios in Vietnamese, AI auto-creates flow

Auto Tool Binding

Auto-detect and connect required API/database

Autonomous Decisions

Agent decides escalation based on context

Continuous Learning

Improve from feedback and real call data

Post-call Analytics

Detailed analytics dashboard for every call — sentiment, intent, agent performance, and business insights

Sentiment AnalysisPositive 78%

Intent Accuracy96.5%

Resolution Rate89%

Avg. Handle Time2m 15s

Auto transcript & recording
Per-sentence sentiment analysis
Script improvement suggestions
Export PDF/CSV reports

Benchmark & Self-hosted Model

STT, TTS, LLM models developed in-house, optimized for Vietnamese with superior accuracy — supports self-hosting on your own infrastructure

Criteria	AntcoAI	VNPT AI	FPT.AI	Google STT
STT Accuracy (Vi) Standard Vietnamese	98.2%	94.5%	95.1%	93.8%
STT Accuracy (Vi dialects) Regional dialects	96.5%	88.2%	89.7%	85.4%
STT Latency (Streaming) First token response	~80ms	~250ms	~200ms	~150ms
TTS Naturalness (MOS) Mean Opinion Score /5	4.6	4.0	4.1	4.3
TTS Latency Time to first audio byte	~60ms	~300ms	~250ms	~180ms
LLM Response (Vi) Vietnamese response quality	9.2/10	7.5/10	7.8/10	8.5/10
LLM Latency Time to first token	~90ms	~400ms	~350ms	~200ms
End-to-end Latency STT → LLM → TTS pipeline	~280ms	~1200ms	~900ms	~600ms
Self-hosted Support On-premise deployment	Full	Limited	Limited	N/A
Vietnamese Voices Number of voices	20+	8	10	4
Concurrent Calls Concurrent calls / node	500+	100	150	200

* Benchmark performed on VIVOS & VLSP 2023 datasets for Vietnamese. Results measured on same hardware (NVIDIA A100 80GB). Updated Q1/2026.

Self-hosted Model

Deploy on your own infrastructure

All STT, TTS and LLM models are developed in-house by AntcoAI team, supporting on-premise or private cloud deployment. Data never leaves your infrastructure — ensuring compliance with data security regulations.

On-premise

Deploy on your own servers

Private Cloud

AWS/Azure/GCP private

Air-gapped

Internal network, no internet

Custom Fine-tune

Train on your own data

Optimized model size: STT 800MB, TTS 1.2GB, LLM 7B params
Minimum requirement: 1x NVIDIA T4 16GB or equivalent
Docker & Kubernetes deployment ready
GPU inference support (CUDA, TensorRT)
Auto-scaling by concurrent calls
Monitoring & alerting with Prometheus/Grafana

Latency Benchmark (ms)

Component	P50	P90	P99
STT (Streaming) First partial result	45ms	78ms	120ms
STT (Final) Complete transcription	80ms	135ms	210ms
NLU Intent Intent classification	12ms	25ms	45ms
LLM Response First token generation	65ms	95ms	150ms
LLM Complete Full response generated	180ms	320ms	500ms
TTS Synthesis First audio chunk	35ms	60ms	95ms
TTS Complete 10s audio generated	85ms	140ms	220ms
End-to-end User speaks → Bot replies	280ms	420ms	650ms

* Benchmark on NVIDIA A100 80GB, batch size 1, Vietnamese. P50/P90/P99 = percentile latency.

Docker Deployment

docker run -d --gpus all \
  -p 8080:8080 \
  -e LICENSE_KEY=YOUR_LICENSE_KEY \
  -v /data/models:/models \
  antcoai/voice-bot:latest \
  --stt-model /models/stt-vi-v3 \
  --tts-model /models/tts-vi-v3 \
  --llm-model /models/llm-vi-7b \
  --max-concurrent 500

SDK for Desktop, Web & Mobile

Integrate Voice Bot into any platform with just a few lines of code — Web, React Native, iOS, Android, and Desktop (Electron)

Web SDK

Integrate Voice Bot directly into websites with a voice call widget, supporting WebRTC

WebRTC voice call

Floating widget UI

Custom branding

Event callbacks

Common SDK Features

Real-time voice call via WebRTC / native audio
TTS & STT support with low latency (<100ms)
Fully customizable UI widget
Event-driven architecture with callbacks
Auto-reconnect on connection loss
Multi-language support (Vietnamese, English, Japanese, Korean)

Web SDK

import { AntcoVoiceBot } from '@antcoai/voice-web-sdk';

const bot = new AntcoVoiceBot({
  apiKey: 'YOUR_API_KEY',
  botId: 'your-bot-id',
  language: 'vi-VN',
  position: 'bottom-right'
});

bot.on('message', (msg) => {
  console.log('Bot:', msg.text);
});

bot.mount('#voice-widget');

Real-world Applications

Voice Bot AI has been successfully deployed at many large enterprises across various industries

Banking & Finance

Voice-based transaction authentication
Balance & transaction history lookup
Credit card payment reminders
Financial product consultation

E-commerce

Order status tracking
Product return support
Size & color consultation
VIP customer care

Healthcare

Auto appointment scheduling
Follow-up & medication reminders
Preliminary symptom consultation
Test result notifications

Real Estate

Project & pricing consultation
Model house visit scheduling
Project progress updates
Installment payment reminders

Logistics & Shipping

Voice-based tracking lookup
Delivery success notifications
Lost package complaint handling
Delivery address confirmation

Education

Course & learning path consultation
Class schedule & assignment reminders
Foreign language pronunciation assessment
Enrollment support

Voice AI Playground

Explore and test Voice AI features with an intuitive interface

Tap to start

Tap the button to start a conversation

Conversation

Press the record button to start a conversation...

Speech Recognition

98% accuracy STT

Natural Language

Context & intent understanding

Voice Synthesis

Natural-sounding TTS

Real-time

Processing under 100ms

Ready to Deploy Voice Bot?

Contact Now