Sales Audio Call Analyzer

Extract actionable insights from sales conversations through automated transcription and AI-powered analysis. Support for multiple input formats including direct audio upload, meeting platform URLs, and text transcripts.

Features

Multi-format Input: Audio files, Fathom/Zoom URLs, or plain text transcripts
Automated Transcription: Speech-to-text conversion using Gemini Files API
Context Enhancement: Optional document and website content integration
Sales-focused Analysis: Specialized prompts for objection handling, sentiment, and action items
Real-time Processing: Streaming analysis results with cost tracking

Architecture Overview

The analyzer supports three distinct processing modes based on input type:

Audio Upload → Transcription → Analysis → Insights
Meeting URL → Extraction → Analysis → Insights  
Text Input → Direct Analysis → Insights

Processing Modes

Audio Service

Processes uploaded audio files through transcription and analysis. Supported Formats: MP3, WAV, M4A, FLAC
File Size Limit: 25MB
Duration Limit: 2 hours

Phantom Service

Extracts transcripts from meeting platform URLs using browser automation. Supported Platforms: Fathom, Zoom, Google Meet
Method: Playwright headless browser extraction
Authentication: Platform-specific session handling

Transcript Service

Analyzes pre-existing text transcripts directly. Input Format: Plain text
Size Limit: 50,000 characters
Encoding: UTF-8

Implementation

Audio Processing Workflow

def process_audio_call(audio_file, supplemental_data=None):
    # 1. Initialize components
    llm = initialize_llm()
    
    # 2. Upload and transcribe
    file_response = gemini_files_api.upload(audio_file)
    transcript = gemini_files_api.transcribe(file_response)
    
    # 3. Process supplemental content
    context = ""
    if supplemental_data:
        if supplemental_data.get('url'):
            context += scrape_website(supplemental_data['url'])
        if supplemental_data.get('documents'):
            context += extract_document_text(supplemental_data['documents'])
    
    # 4. Analyze conversation
    analysis_chain = create_sales_analysis_chain(llm)
    insights = analysis_chain.process(transcript, context)
    
    # 5. Cleanup and logging
    cleanup_temp_files(audio_file)
    cost = calculate_token_cost(insights.token_usage)
    store_analysis_results(insights, cost)
    
    return insights

URL Extraction Workflow

def process_meeting_url(meeting_url, supplemental_data=None):
    # 1. Initialize browser automation
    playwright_session = initialize_playwright()
    
    # 2. Extract transcript
    transcript = playwright_session.extract_transcript(meeting_url)
    
    # 3. Process supplemental content (same as audio service)
    context = process_supplemental_data(supplemental_data)
    
    # 4. Analyze and return results
    return analyze_sales_conversation(transcript, context)

API Reference

Analysis Endpoint

POST /api/sales-call-analysis

Request Body (Audio Upload)

{
  "inputType": "audio",
  "audioFile": "base64_encoded_audio_data",
  "fileName": "sales_call.mp3",
  "supplementalData": {
    "websiteUrl": "https://client-company.com",
    "documents": ["proposal.pdf", "pricing.docx"]
  },
  "analysisOptions": {
    "focusAreas": ["objections", "sentiment", "next_steps"],
    "generateSummary": true
  }
}

Request Body (Meeting URL)

{
  "inputType": "url",
  "meetingUrl": "https://fathom.video/call/abc123",
  "platformType": "fathom",
  "supplementalData": {
    "websiteUrl": "https://client-company.com"
  }
}

Request Body (Text Transcript)

{
  "inputType": "transcript",
  "transcriptText": "Sales rep: Thanks for joining today's call...",
  "supplementalData": {
    "documents": ["product_spec.pdf"]
  }
}

Response Format

{
  "analysisId": "sca_12345",
  "status": "completed",
  "insights": {
    "summary": "Client showed strong interest in automation features",
    "sentiment": {
      "overall": "positive",
      "score": 0.75,
      "keyMoments": [
        {
          "timestamp": "00:05:23",
          "sentiment": "concern",
          "topic": "implementation timeline"
        }
      ]
    },
    "objections": [
      {
        "objection": "Worried about integration complexity",
        "response": "Provided case study of similar client success",
        "resolved": true
      }
    ],
    "actionItems": [
      "Send technical integration guide",
      "Schedule follow-up demo for next Tuesday"
    ],
    "talkRatio": {
      "salesRep": 45,
      "client": 55
    }
  },
  "metadata": {
    "processingTime": "23s",
    "tokenUsage": 3200,
    "cost": 0.048,
    "transcriptLength": 1450
  }
}

Core Components

Component	Purpose
Gemini Files API	Audio upload and transcription
Playwright Engine	Browser automation for URL extraction
Web Scraper	Website content extraction
Document Parser	PDF, DOCX, TXT content extraction
LLM Analysis Chain	Sales-focused conversation analysis
Prompt Templates	Structured analysis instructions
Cost Tracker	Token usage and pricing calculation

Configuration

Audio Processing Settings

{
  "transcription": {
    "model": "gemini-pro",
    "language": "auto-detect",
    "speakerDiarization": true,
    "maxDuration": 7200
  },
  "fileValidation": {
    "maxSize": "25MB",
    "allowedFormats": ["mp3", "wav", "m4a", "flac"]
  }
}

Analysis Parameters

{
  "analysis": {
    "focusAreas": [
      "objection_handling",
      "sentiment_analysis", 
      "talk_ratio",
      "action_items",
      "decision_makers"
    ],
    "summaryLength": "detailed",
    "includeTimestamps": true
  }
}

Usage Examples

Audio File Analysis

curl -X POST /api/sales-call-analysis \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer your-api-key" \
  -d '{
    "inputType": "audio",
    "audioFile": "data:audio/mp3;base64,UklGRnoG...",
    "fileName": "client_call.mp3",
    "analysisOptions": {
      "focusAreas": ["objections", "next_steps"]
    }
  }'

Meeting URL Analysis

curl -X POST /api/sales-call-analysis \
  -H "Content-Type: application/json" \
  -d '{
    "inputType": "url",
    "meetingUrl": "https://fathom.video/call/xyz789",
    "platformType": "fathom",
    "supplementalData": {
      "websiteUrl": "https://prospect-company.com"
    }
  }'

Python Implementation

import requests
import base64

# Audio file analysis
with open('sales_call.mp3', 'rb') as audio_file:
    audio_data = base64.b64encode(audio_file.read()).decode()

response = requests.post('/api/sales-call-analysis', json={
    'inputType': 'audio',
    'audioFile': f'data:audio/mp3;base64,{audio_data}',
    'fileName': 'sales_call.mp3',
    'analysisOptions': {
        'focusAreas': ['objections', 'sentiment', 'action_items'],
        'generateSummary': True
    }
})

analysis = response.json()
print(f"Summary: {analysis['insights']['summary']}")
print(f"Action Items: {analysis['insights']['actionItems']}")

Performance & Limits

Processing Metrics

Audio Transcription: 1-3 minutes per hour of audio
URL Extraction: 15-45 seconds depending on platform
Text Analysis: 5-15 seconds for typical call transcripts
Concurrent Processing: Up to 10 analyses simultaneously

Rate Limits

API Requests: 100 analyses per hour per API key
File Uploads: 50MB total per hour
Token Usage: Tracked and reported per analysis

Platform Limitations

Fathom: Requires public or shared call links
Zoom: Cloud recordings only, local recordings not supported
Google Meet: Requires Google Workspace with recording enabled

Error Handling

Common Error Responses

Status Code	Error Type	Description	Solution
`400`	`INVALID_AUDIO_FORMAT`	Unsupported audio file format	Convert to MP3, WAV, M4A, or FLAC
`413`	`FILE_TOO_LARGE`	Audio file exceeds size limit	Compress audio or split into segments
`422`	`TRANSCRIPTION_FAILED`	Unable to process audio content	Check audio quality and format
`404`	`URL_NOT_ACCESSIBLE`	Meeting URL is private or invalid	Verify URL is public and accessible
`429`	`RATE_LIMIT_EXCEEDED`	Too many concurrent requests	Implement request throttling

Error Response Format

{
  "error": {
    "code": "TRANSCRIPTION_FAILED",
    "message": "Unable to transcribe audio content",
    "details": {
      "fileName": "sales_call.mp3",
      "fileSize": "15.2MB",
      "duration": "45:23",
      "reason": "Audio quality too low for reliable transcription"
    }
  }
}

Integration Guide

Authentication

All requests require Bearer token authentication:

curl -H "Authorization: Bearer your-api-key"

Webhook Support

Configure webhooks for analysis completion:

{
  "webhookUrl": "https://your-app.com/sales-analysis-complete",
  "events": ["analysis.completed", "analysis.failed", "transcription.completed"]
}

Streaming Results

For real-time progress updates:

GET /api/sales-call-analysis/{analysisId}/stream

Returns Server-Sent Events:

data: {"type": "progress", "step": "transcription", "completion": 30}
data: {"type": "insights", "category": "objections", "data": [...]}
data: {"type": "complete", "analysisId": "sca_12345"}

Overview

Setup

Your Account

Features

Policies

Sales Audio Call Analyzer

Features

Architecture Overview

Processing Modes

Audio Service

Phantom Service

Transcript Service

Implementation

Audio Processing Workflow

URL Extraction Workflow

API Reference

Analysis Endpoint

Core Components

Configuration

Audio Processing Settings

Analysis Parameters

Usage Examples

Audio File Analysis

Meeting URL Analysis

Python Implementation

Performance & Limits

Processing Metrics

Rate Limits

Platform Limitations

Error Handling

Common Error Responses

Error Response Format

Integration Guide

Authentication

Webhook Support

Streaming Results

Overview

Setup

Your Account

Features

Policies

​Features

​Architecture Overview

​Processing Modes

​Audio Service

​Phantom Service

​Transcript Service

​Implementation

​Audio Processing Workflow

​URL Extraction Workflow

​API Reference

​Analysis Endpoint

​Core Components

​Configuration

​Audio Processing Settings

​Analysis Parameters

​Usage Examples

​Audio File Analysis

​Meeting URL Analysis

​Python Implementation

​Performance & Limits

​Processing Metrics

​Rate Limits

​Platform Limitations

​Error Handling

​Common Error Responses

​Error Response Format

​Integration Guide

​Authentication

​Webhook Support

​Streaming Results

Features

Architecture Overview

Processing Modes

Audio Service

Phantom Service

Transcript Service

Implementation

Audio Processing Workflow

URL Extraction Workflow

API Reference

Analysis Endpoint

Core Components

Configuration

Audio Processing Settings

Analysis Parameters

Usage Examples

Audio File Analysis

Meeting URL Analysis

Python Implementation

Performance & Limits

Processing Metrics

Rate Limits

Platform Limitations

Error Handling

Common Error Responses

Error Response Format

Integration Guide

Authentication

Webhook Support

Streaming Results