API Documentation

Get up and running in under five minutes. Install the SDK for your programming language, retrieve your API credentials, and initiate your first audio transcription request.

pip install verbalyze
# or
npm install @verbalyze/sdk

import verbalyze as vb

client = vb.Client(api_key="vb_sk_...")

# Transcribe a Hindi audio call
result = client.transcribe(
    audio="call_recording.wav",
    language="hi-IN",
    domain="banking" # fine-tuned for financial contexts
)

print(result.text)
# → "नमस्ते, मेरा खाता बंद हो गया है"
print(f"Confidence: {result.confidence} | Latency: {result.latency_ms}ms")

All requests to the Verbalyze API must be authenticated using your account credentials. Place your token inside the HTTP `Authorization` header prefixed with `Bearer`.

Authorization: Bearer vb_sk_your_api_key_here
Content-Type: application/json

Verbalyze STT converts spoken words in Indian accents and local dialects into accurate, high-fidelity text. We support two primary methods:

Synthesize natural, human-like voice response output from structured scripts.

{
  "text": "आपका स्वागत है, मैं आपकी किस प्रकार सहायता कर सकता हूँ?",
  "voice_id": "hi-IN-Deepak",
  "speed": 1.05,
  "output_format": "mp3"
}

Program complex state machines and natural conversational trees that run on outbound calls.

{
  "agent_id": "agent_bfsi_emi_v3",
  "recipient": "+919876543210",
  "variables": {
    "customer_name": "Rajesh Kumar",
    "due_amount": "₹4,500",
    "due_date": "July 1st"
  }
}

Receive real-time push event logs to your webhook URL whenever a call status changes or updates.

{
  "event": "call.completed",
  "call_id": "call_982341829",
  "duration_seconds": 45,
  "sentiment": "positive",
  "outcome": "promise_to_pay"
}

Verbalyze operates on top of rigorous security frameworks, ensuring your organization aligns with regulatory rules in India.