🎤 Voice Input for CLARISSA¶

Voice input enables hands-free interaction with CLARISSA - speak your reservoir engineering questions naturally.

Quick Start¶

Open the demo in Chrome/Edge
Enter your OpenAI API key
Click 🎤 or use keyboard shortcut
Speak your question
See real-time transcription

Resources¶

Resource	Description
Voice Demo	Interactive browser demo
Tutorial Guide	Step-by-step walkthrough
Colab Notebook	Jupyter notebook with examples
ADR-028	Architecture decision record

Recording Demo Videos¶

Need to record a demo? Multiple options:

Browser-Based (Zero Install)¶

Screen Recorder - Basic recording
Screen Recorder + PiP - With webcam overlay

Command Line Tools (macOS/Linux)¶

# Setup (one time)
git clone https://gitlab.com/wolfram_laube/blauweiss_llc/irena.git ~/Projects/clarissa
~/Projects/clarissa/tools/recording/setup.sh

# Use (after restart or source ~/.zshrc)
record-pip.sh start 60     # 60s with camera PiP
record-timed.sh 30         # 30s screen only
record-demo.sh             # Interactive menu

Structure:

tools/recording/
├── setup.sh              # Auto-detects macOS/Linux
├── common/config.sh      # Shared settings
├── macos/                # macOS-specific scripts
│   ├── record-pip.sh
│   ├── record-timed.sh
│   └── record-applescript.sh
└── linux/                # Linux-specific scripts
    ├── record-pip.sh
    └── record-timed.sh

📋 Showcase Guide: See the Demo Showcase Guide for step-by-step instructions on what to say, demonstrate, and expect in your demo video.

See Recording Tools Documentation for full details.

Implementation Status¶

Component	Status	Notes
Browser Demo	✅ Live	WebSpeech API
Tutorial Guide	✅ Complete	Step-by-step
Colab Notebook	✅ Complete	With mic workaround
Recording Tools	✅ Complete	macOS + Linux
ADR-028	✅ Documented	Architecture
Phase 1 Backend	🔜 Planned	Issues #67-72

Architecture¶

Voice input uses the WebSpeech API for browser-based transcription:

User Speech → Microphone → WebSpeech API → Text → CLARISSA

For production deployment, the architecture supports multiple backends: - OpenAI Whisper (cloud) - Local Whisper (privacy-focused) - WebSpeech (browser-native)

See ADR-028 for details.

#66 Voice Input Feature - Epic
#67-72 - Implementation tasks