Version 2 — refreshed architecture with multi-vendor model discovery, local Whisper transcription, and improved multimodal workflows.

Author: Arpit J Soni
LinkedIn: linkedin.com/in/arpit-j-soni

Download the Windows installer: Download the Windows installer

Product Hunt: Product Hunt

Prerequisites

Required Software

• Node.js 18+ and npm 9+
• Python 3.8+ (for local Whisper transcription)
• macOS 13+ or Windows 10/11

API Keys (Optional but Recommended)

• OpenAI API key (for ChatGPT, GPT-4, etc.) - Set as OPENAI_API_KEY or REACT_APP_OPENAI_API_KEY
• Google Gemini API key (for Gemini models)
• Anthropic API key (for Claude models)
• Perplexity API key (for Perplexity models)

Note: The app works with local Whisper transcription without any API keys, but API keys are required for AI responses and meeting summaries/questions.

Installation & Setup

Step 1: Install Node.js Dependencies

# From project root
npm ci

Step 2: Install Python Dependencies (for Local Transcription)

The app uses a local Whisper server for real-time speech-to-text transcription. Install Python dependencies:

# Install Python packages directly
pip install faster-whisper websockets numpy torch

# Or if you need to use pip3:
pip3 install faster-whisper websockets numpy torch

# Or install with specific versions (recommended)
pip install "faster-whisper>=1.0.0" "websockets>=12.0" "numpy>=1.24.0" "torch>=2.0.0"

Required Python packages:

• faster-whisper>=1.0.0 - Optimized Whisper implementation for real-time transcription
• websockets>=12.0 - WebSocket support for real-time streaming (optional, for future features)
• numpy>=1.24.0 - Numerical computing library
• torch>=2.0.0 - PyTorch for model inference

Note: On first run, the Whisper small.en model (~500-600MB) will be automatically downloaded to whisper_models/ directory.

Step 3: Run the Application

# Start the React dev server and Electron app
npm run electron-dev

This command:

• Starts the React development server on http://localhost:3000
• Automatically starts the local Whisper server on port 8765
• Launches the Electron app once the React server is ready

Step 4: Configure API Keys (Optional)

1. Open the app and click ⚙ Settings
2. Select your preferred vendor (OpenAI, Gemini, Anthropic, or Perplexity)
3. Enter your API key in the settings panel
4. Save the configuration

Version 2 Highlights

• Choose your favorite LLM for every request—swap between OpenAI (ChatGPT), Google Gemini, Perplexity, or Claude (Anthropic) right from the app.
• Local Whisper Transcription - Real-time speech-to-text using local Whisper model (no API calls needed for transcription).
• Unified multimodal support across all providers: mix text plus multiple screenshots in one prompt and get a single streaming answer.
• Smarter fallbacks so retired or unavailable models are replaced automatically without you touching settings.
• Cleaner setup that keeps your provider configuration in sync and ready for future upgrades (voice, tools, etc.).

Core Features

Main Interface

• Screenshot capture with automatic window-hide during capture (restores immediately).
• Multimodal querying (text + screenshots) with streaming AI responses.
• Audio recording with automatic transcription (supports vendors with audio capabilities).
• Conversation history with sidebar panel for reviewing past interactions.
• Response window with markdown rendering and syntax highlighting.

Meetings Page

• Real-time local transcription using Whisper (no API required for transcription).
• Dual audio capture:
  • Microphone capture - Records your voice
  • System audio capture - Records system/application audio
• Independent controls:
  • Start/Stop Microphone independently
  • Start/Stop System Audio independently
• Live transcription display with separate streams for microphone and system audio.
• AI-powered meeting summary - Auto-generates summaries every 5 seconds (requires API key).
• Question extraction - Automatically identifies and extracts questions from transcript every 10 seconds (requires API key).
• Export functionality - Export transcriptions, summaries, and questions to text file.

Privacy & Window Management

• Screen-share privacy: Hide app from screen sharing while keeping it visible for you.
• Draggable windows: Drag the window by clicking and dragging the header area.
• Window opacity control: Adjust app transparency for overlay use.
• Always-on-top: Window stays on top of other applications.

Settings & Customization

• Multi-vendor support: Switch between OpenAI, Gemini, Anthropic, and Perplexity.
• Theme selection: Light, Dark, or Auto (system preference).
• Font size control: Small, Medium, or Large.
• Customizable keyboard shortcuts for all actions.
• Participant names: Set custom names for microphone and system audio in meetings.
• Screenshot limits: Configure maximum number of screenshots per session.

Theming

• Aurora-inspired palette with CSS variables applied app-wide.
• Smooth transitions and modern UI components.
• Accessibility support: High contrast mode and reduced motion preferences.

Local Whisper Transcription

The app includes a built-in local Whisper server for real-time speech-to-text transcription:

• Automatic startup: Whisper server starts automatically when the app launches.
• No API required: Transcription works completely offline (no API calls needed).
• Low latency: Optimized for real-time transcription with 5-second audio chunks.
• Model: Uses faster-whisper with small.en model for English transcription.
• Server endpoint: Runs on http://localhost:8765

Manual Whisper Server Control

If you need to start/stop the Whisper server manually:

# Start server manually
python3 whisper_server.py

# Or with custom port
WHISPER_PORT=8765 python3 whisper_server.py

The server provides:

• GET /health - Health check endpoint
• POST /transcribe - Audio transcription endpoint

Build Desktop App

macOS

Requirements:

• Xcode Command Line Tools
• Apple Developer account for signing/notarization (recommended for distribution)

Build steps:

npm ci
npm run electron-pack

Artifacts:

• dist/Visual Studio Code Desktop-<version>-arm64.dmg (and .zip)
• Unpacked .app under dist/mac-arm64/

Optional signing & notarization:

export CSC_IDENTITY_AUTO_DISCOVERY=true
export APPLE_ID="your-apple-id@example.com"
export APPLE_APP_SPECIFIC_PASSWORD="xxxx-xxxx-xxxx-xxxx"
npm run electron-pack

Windows

Requirements:

• Windows 10/11
• Optional code-signing certificate for best install UX

Build steps:

npm ci
npm run electron-pack

Artifacts:

• NSIS installer: dist/Visual Studio Code Desktop-<version>.exe

Optional Windows signing (via electron-builder):

• Set CSC_LINK (path/base64 to PFX) and CSC_KEY_PASSWORD, or use store-based options.

Keyboard Shortcuts

In-App

• CmdOrCtrl+H: Take screenshot (also registered globally)
• CmdOrCtrl+R: Reset the current session
• CmdOrCtrl+S: Save the current response
• Shift+Enter (while holding Cmd/Ctrl): Submit the current query

Global

• CmdOrCtrl+Shift+V: Toggle "visible to screen share" (privacy mode on/off; window stays visible to you)
• CmdOrCtrl+Up/Down/Left/Right: Move window by 50px
• CmdOrCtrl+Alt+. / CmdOrCtrl+Alt+,: Increase / Decrease app opacity

Note: All keyboard shortcuts can be customized in Settings.

Troubleshooting

Whisper Server Issues

Server not starting:

• Verify Python 3.8+ is installed: python3 --version
• Check dependencies: pip3 list | grep faster-whisper
• Check if port 8765 is available: lsof -i :8765 (macOS) or netstat -ano | findstr :8765 (Windows)
• Review server logs in console or /tmp/whisper_server.log

Transcription not working:

• Ensure Whisper server is running: Check http://localhost:8765/health
• Verify microphone permissions are granted
• Check browser console for connection errors
• Restart the Electron app to restart the Whisper server

Model download issues:

• First-time model download requires internet connection
• Model is saved to whisper_models/ directory (~500-600MB)
• If download fails, delete whisper_models/ and restart the app

General Issues

Python not found:

• macOS/Linux: Install Python 3 using your package manager or from python.org
• Windows: Download from python.org, ensure "Add Python to PATH" is checked during installation

Dependencies installation fails:

• Try upgrading pip: pip install --upgrade pip or pip3 install --upgrade pip
• Use virtual environment: python3 -m venv venv && source venv/bin/activate (macOS/Linux) or python -m venv venv && venv\Scripts\activate (Windows)

Buttons not working after returning to Meetings page:

• The app automatically re-initializes event listeners when returning to the page
• If issues persist, restart the Electron app

Window not draggable:

• Drag the window by clicking and holding on the header area (not on buttons)
• Ensure you're not clicking on interactive elements (buttons, inputs, etc.)

Generate Icons

This repo exposes a helper command that forwards flags to electron-icon-maker.

Source image: A square PNG with sufficient resolution (at least 1024×1024 recommended).

Command:

npm run make-icons -- --input /absolute/path/to/source.png --output /absolute/path/to/output-dir

Example using this repo's asset layout:

npm run make-icons -- \
  --input Desktop/CHEAT/CHEATS/public/assets/vscode.png \
  --output Desktop/CHEAT/CHEATS/public/assets

The generated .icns (mac) and .ico (windows) can be referenced by the build config and at runtime (tray/window icons).

Change the Application Name

Update in package.json:

• name: npm package name (optional for distribution but keep consistent)
• build.productName: The user-facing application name in installers and the .app/.exe
• build.appId: Reverse-DNS identifier (change to your domain, e.g., com.yourco.product)

Optional UI title: If you want a custom window title beyond the product name, set it via the renderer or in BrowserWindow options.

Change the Application Icon

Build-time (installers):

• macOS: build.mac.icon → public/assets/your.icns
• Windows: build.win.icon → public/assets/your.ico

Runtime (tray/window):

• File: public/electron.js picks icons per platform from public/assets (vscode.icns on mac, vscode.ico on win, vscode.png as fallback for others). Replace those file paths with your own.

After replacing icons, rebuild:

npm run electron-pack

Social Links

• Personal site: arpitjsoni.com
• Twitter/X: x.com/arpitsoni1893
• YouTube: youtube.com/@ArpitJSoni_1

References

• Twitter/X profile mentioned above for author identity: https://x.com/arpitsoni1893