Skip to content

Latest commit

 

History

History
702 lines (480 loc) · 16 KB

README.md

File metadata and controls

702 lines (480 loc) · 16 KB

Transcript Parser - User Manual

Table of Contents

  1. Overview
  2. Getting Started
  3. Core Features
  4. Authentication & API Configuration
  5. Video Processing
  6. Transcript Management
  7. AI-Powered Features
  8. Cost Tracking & Billing
  9. Export Options
  10. Keyboard Shortcuts
  11. Advanced Features
  12. Troubleshooting

Overview

Transcript Parser is an advanced AI-powered desktop application that converts video/audio files into searchable, editable transcripts with speaker diarization. Built with Electron, React, and Google Gemini AI, it provides professional-grade transcription with intelligent speaker identification and comprehensive editing capabilities.

Key Capabilities

  • AI-Powered Transcription: Uses Google Gemini 2.5 Flash for accurate speech-to-text
  • Speaker Diarization: Automatically identifies and separates different speakers
  • AI Name Detection: Intelligently detects speaker names from introductions
  • Real-Time Editing: Edit transcripts with full undo/redo support
  • Advanced Search: Search across transcripts with highlighting
  • Cost Tracking: Real-time token usage and monthly billing breakdown
  • Multiple Export Formats: TXT, JSON, SRT, VTT formats
  • Cross-Platform: Available for Windows, macOS, and Linux

Getting Started

Installation

Windows

  1. Download Transcript Parser-Setup-1.0.0.exe from the releases
  2. Run the installer and follow the setup wizard
  3. Launch Transcript Parser from Start Menu

Portable Version

  • Download Transcript Parser-Portable-1.0.0.exe
  • Run directly without installation (no admin rights required)

First Launch

  1. Set up API Access

    • Click the settings icon (⚙️) in the top-right
    • Choose your API configuration method
    • Enter credentials as needed

  2. Load Your First Video

    • Click "Choose Video File" or drag & drop a video
    • Supported formats: MP4, AVI, MOV, WebM, MP3, WAV
    • Maximum file size: 2GB recommended

Core Features

1. Video/Audio Upload

Upload Methods:

  • Click Button: Click "Choose Video File" button
  • Drag & Drop: Drag video/audio files directly into the upload area
  • Recent Files: Access recently processed files from history

Supported Formats:

  • Video: .mp4, .avi, .mov, .webm
  • Audio: .mp3, .wav, .m4a, .flac

Processing:

  • File size: Up to 2GB (recommended)
  • Conversion: Automatically converts to WebM/Opus for optimal processing
  • Progress: Real-time progress bar with percentage and status updates

2. Automatic Transcription

How It Works:

  1. Upload your media file
  2. Click "Start Transcription"
  3. AI processes audio and generates transcript
  4. Speakers are automatically identified and separated

Transcription Features:

  • Automatic Speaker Detection: Identifies unique speakers
  • Timestamps: Precise start/end times for each segment
  • Confidence Scores: Quality indicators for each segment
  • Real-Time Updates: See transcript build as processing completes

Authentication & API Configuration

Configuration Modes

1. Own API Key (Recommended)

Best for: Individual users, full control

  1. Click Settings (⚙️) → "API Configuration"
  2. Select "Use Own API Key"
  3. Get API key from Google AI Studio
  4. Paste key and click "Save"

Advantages:

  • Full cost control
  • No access codes needed
  • Direct billing through Google

2. Access Code

Best for: Shared access, organizations

  1. Select "Use Access Code"
  2. Enter 10-digit code: XXX-XXXX-XXX
  3. Developer's API key is used automatically

Access Code Validation:

  • Format: 3-4-3 digits
  • Example: 123-4567-890
  • Contact admin for codes

3. Paid Service (Future)

Coming soon: Pay-as-you-go with monthly billing

Video Processing

Video Player Features

Playback Controls:

  • ▶️ Play/Pause: Click video or press Space
  • ⏩ Fast Forward: Press → or L
  • ⏪ Rewind: Press ← or J
  • 🔇 Mute/Unmute: Press M
  • 📽️ Fullscreen: Press F

Transcript Sync:

  • Click Entry: Jump to timestamp in video
  • Auto-Highlight: Current speaking segment highlighted
  • Seek Bar: Visual timeline with segment markers

Advanced Controls:

  • Playback speed: 0.25x to 2x
  • Volume control: 0% to 100%
  • Frame-by-frame navigation

Transcript Management

Viewing Transcripts

Layout:

  • Left Panel: Speaker analytics and statistics
  • Center Panel: Transcript entries with search
  • Right Panel: (Optional) Video player

Entry Information:

  • Speaker name/color
  • Start and end timestamps
  • Confidence percentage
  • Full text content

Editing Transcripts

Enable Editing:

  • Toggle edit mode in top toolbar
  • Double-click any entry to edit
  • Modify text, start time, or end time

Edit Operations:

  1. Text Editing:

    • Double-click entry text
    • Make changes in textarea
    • Click "Save" or press Enter

  2. Timestamp Editing:

    • Click edit icon
    • Modify start/end times (in seconds)
    • Format: Decimal (e.g., 12.5)

  3. Undo/Redo:

    • Undo: Ctrl+Z (Windows) / Cmd+Z (Mac)
    • Redo: Ctrl+Shift+Z / Cmd+Shift+Z

Visual Indicators:

  • 🔄 Edited Badge: Shows modified entries
  • ✓ Save Confirmation: Visual feedback on save
  • ❌ Cancel Option: Discard changes

Speaker Management

Rename Speakers:

  1. Find speaker in left panel
  2. Click edit icon (✏️) next to speaker name
  3. Type new name
  4. Press Enter or click ✓
  5. Name updates across all entries

Speaker Colors:

  • Blue, Emerald, Purple, Orange, Pink, Cyan
  • Automatically assigned
  • Consistent throughout transcript

AI-Powered Features

AI Speaker Name Detection

Automatic Detection:

  1. Click "Detect Names" button (✨ Sparkles icon)
  2. AI analyzes first 30 entries of each speaker
  3. Looks for self-introduction patterns
  4. Returns suggestions with confidence levels

Detection Patterns:

  • "My name is [name]"
  • "I'm [name]"
  • "This is [name]"
  • "Hi, I'm [name]"
  • "[name] here"

Confidence Levels:

  • High: Clear, unambiguous introduction (e.g., "My name is John Smith")
  • Medium: Less formal introduction (e.g., "I'm Sarah")
  • Low: Ambiguous or indirect reference

Review Suggestions:

  • Evidence Quote: See exact text where name was detected
  • Accept: Applies name to all speaker entries
  • Reject: Dismisses suggestion
  • Dismiss All: Remove all suggestions

Search & Filter

Text Search:

  1. Type query in search box
  2. Live results with match count
  3. Highlighting in transcript entries
  4. Case-insensitive matching

Filters:

  • Speaker Filter: Show only specific speakers
  • Time Range: Filter by start/end timestamps
  • Confidence: (Future) Filter by confidence score

Combined Filters:

  • Search + Speaker: Find text from specific speaker
  • Search + Time: Find text in time range
  • All filters stack together

Cost Tracking & Billing

Real-Time Cost Tracking

What's Tracked:

  • Input tokens (prompt + audio analysis)
  • Output tokens (generated transcript)
  • Total tokens per operation
  • Estimated cost in USD

Cost Calculation:

  • Gemini 2.5 Flash: $0.075/1M input, $0.30/1M output
  • Real-time updates from API responses
  • Persists across sessions (localStorage)

Viewing Cost Summary

  1. Click "Cost Summary" button (💰)
  2. See overview cards:
    • Total Tokens Used
    • Total Cost (USD)
    • Total Operations
    • Average Cost per Operation

Monthly Billing Breakdown

Current Month Card:

  • Tokens used this month
  • Current month cost
  • Operations count
  • Highlighted in amber/gold

Historical Billing:

  • All past months sorted newest first
  • Monthly totals: tokens, cost, operations
  • Format: "December 2024"

Usage by Category:

  • By Model: Gemini 2.5 Flash, 1.5 Flash, etc.
  • By Operation: Video Transcription, Name Detection

Export Options

Export Formats

1. Plain Text (.txt)

  • Speaker labels
  • Timestamps in [HH:MM:SS] format
  • Clean, readable format
  • Best for: Documentation, notes

Example:

[00:00:05] Speaker 1: Hello everyone, welcome to the meeting.
[00:00:12] Speaker 2: Thanks for having me.

2. JSON (.json)

  • Complete structured data
  • All metadata preserved
  • Speakers, timestamps, confidence
  • Best for: Developers, data analysis

Example:

{
  "entries": [{
    "id": "1",
    "speaker": "Speaker 1",
    "speakerNumber": 1,
    "startTime": 0.0,
    "endTime": 5.2,
    "text": "Hello everyone",
    "confidence": 0.95
  }],
  "speakers": [...],
  "metadata": {...}
}

3. SubRip (.srt)

  • Standard subtitle format
  • Numbered sequences
  • Timestamp format: HH:MM:SS,mmm
  • Best for: Video subtitles, YouTube

Example:

1
00:00:00,000 --> 00:00:05,200
Speaker 1: Hello everyone

2
00:00:05,200 --> 00:00:12,000
Speaker 2: Thanks for having me

4. WebVTT (.vtt)

  • Web video text tracks
  • HTML5 compatible
  • Metadata support
  • Best for: Web players, accessibility

Example:

WEBVTT

00:00:00.000 --> 00:00:05.200
<v Speaker 1>Hello everyone

00:00:05.200 --> 00:00:12.000
<v Speaker 2>Thanks for having me

Export Process

  1. Click "Export" button (📥)
  2. Select format from dropdown
  3. Choose save location
  4. File is generated and downloaded

Keyboard Shortcuts

Global Shortcuts

Shortcut Action
Ctrl/Cmd + O Open video file
Ctrl/Cmd + S Save transcript
Ctrl/Cmd + E Export transcript
Ctrl/Cmd + F Focus search box
Ctrl/Cmd + Z Undo edit
Ctrl/Cmd + Shift + Z Redo edit
Escape Clear search/filters

Video Player Shortcuts

Shortcut Action
Space Play/Pause
→ or L Skip forward 5s
← or J Skip backward 5s
↑ Volume up
↓ Volume down
M Mute/Unmute
F Toggle fullscreen
0-9 Jump to 0%-90%

Transcript Navigation

Shortcut Action
↑/↓ Navigate entries
Enter Play entry timestamp
Double-Click Edit entry (if enabled)
Ctrl/Cmd + Click Multi-select

Advanced Features

Speaker Analytics

Statistics Displayed:

  • Total speaking time per speaker
  • Percentage of total conversation
  • Number of segments
  • Average segment duration

Visual Indicators:

  • Progress bars for speaking time
  • Color-coded speakers
  • Segment count badges

Transcript History

  • Recently processed files
  • Quick reload previous transcripts
  • Automatic save on process
  • Indexed for fast search

Performance Optimization

Virtual Scrolling:

  • Handles 10,000+ entries smoothly
  • Only renders visible entries
  • Smooth 60fps scrolling

Progressive Loading:

  • Entries load as transcription completes
  • No waiting for full completion
  • Real-time updates

Troubleshooting

Common Issues

1. "API Key Invalid" Error

Solution:

  • Verify API key is correct
  • Check key has Gemini API access enabled
  • Regenerate key from Google AI Studio
  • Ensure billing is enabled on Google Cloud

2. Transcription Fails

Possible Causes:

  • File too large (>2GB)
  • Unsupported format
  • Poor audio quality
  • API quota exceeded

Solutions:

  • Compress video/audio file
  • Convert to supported format (MP4, WebM)
  • Improve audio quality
  • Check API quota limits

3. Video Won't Play

Solutions:

  • Update Electron app to latest version
  • Check video codec compatibility
  • Convert video to WebM format
  • Verify file isn't corrupted

4. Slow Performance

Solutions:

  • Close unnecessary background apps
  • Process smaller files
  • Enable hardware acceleration
  • Increase available RAM

5. Export Fails

Solutions:

  • Check disk space
  • Verify write permissions
  • Choose different save location
  • Check file name validity

Getting Help

Support Channels:

  • GitHub Issues: Report bugs
  • Documentation: Check implementation guides in docs/implementation/
  • Community: (Future) Discord/Slack channels

Before Reporting:

  1. Check this manual
  2. Review error messages
  3. Check console logs (Ctrl+Shift+I in app)
  4. Note steps to reproduce
  5. Include system info (OS, version)

Technical Requirements

Minimum System Requirements:

  • OS: Windows 10+, macOS 10.13+, Ubuntu 18.04+
  • RAM: 4GB (8GB recommended)
  • Disk: 500MB for app + space for videos
  • Internet: Required for transcription API calls

Recommended:

  • RAM: 8GB+ for large files
  • CPU: Multi-core processor for faster processing
  • SSD: For better video loading performance
  • Bandwidth: Stable connection for API calls

Privacy & Security

Data Handling:

  • Local Processing: Videos stay on your device
  • API Transmission: Only audio data sent to Google
  • No Storage: Google doesn't store your audio
  • Encryption: HTTPS for all API calls

API Key Security:

  • Keys stored in localStorage (encrypted)
  • Never transmitted except to Google
  • Rotatable at any time
  • Access codes don't expose developer keys

Best Practices:

  • Don't share API keys
  • Rotate keys periodically
  • Use access codes for teams
  • Review cost usage regularly

Updates & Changelog

Current Version: 1.0.0

Recent Features:

  • ✨ AI-powered speaker name detection
  • 📊 Monthly billing breakdown
  • 💰 Real-time cost tracking
  • 🎨 Enhanced UI with speaker colors
  • ⌨️ Keyboard shortcuts
  • 📝 Inline transcript editing
  • 🔍 Advanced search and filters

Coming Soon:

  • Cloud sync for transcripts
  • Team collaboration features
  • Custom vocabulary/terminology
  • Batch processing multiple files
  • AI summarization
  • Translation support

FAQ

Q: How accurate is the transcription? A: Accuracy depends on audio quality. Gemini 2.5 Flash provides 85-95% accuracy for clear audio with minimal background noise.

Q: Can I edit the transcript? A: Yes! Enable edit mode and double-click any entry to modify text or timestamps.

Q: What languages are supported? A: Currently supports English. Additional languages coming soon via Gemini multilingual models.

Q: How much does transcription cost? A: Cost varies by audio length. Typical 1-hour meeting: ~$0.05-$0.15. Check Cost Summary for exact usage.

Q: Can I process multiple files at once? A: Not currently. Batch processing is planned for a future release.

Q: Is my data private? A: Yes. Videos never leave your device. Only audio is temporarily sent to Google's Gemini API (not stored).

Q: Can I use this offline? A: No. Internet connection required for AI transcription. Local playback and editing work offline.

Q: How long does transcription take? A: Typically 1-3 minutes per hour of audio, depending on network speed and API response time.

License & Credits

License: MIT License Developer: Keven W. Markham AI Model: Google Gemini 2.5 Flash Framework: Electron + React + TypeScript UI Library: shadcn/ui + Tailwind CSS

Third-Party Licenses:

  • React (MIT)
  • Electron (MIT)
  • Google Generative AI SDK (Apache 2.0)
  • Lucide Icons (ISC)

Contact & Support

GitHub: transcript-parser Issues: Report bugs or request features Email: (Contact info) Documentation: See docs/ folder for technical guides

Last Updated: December 19, 2024 • Version 1.0.0