GLaDOS WebUI

A web interface for GLaDOS text-to-speech with AI conversation capabilities. Talk to GLaDOS in your browser and hear her iconic voice respond with characteristic passive-aggressive wit.

Credits

This project is a modified fork of R2D2FISH/glados-tts, the original GLaDOS TTS engine using neural network-based speech synthesis.

What's different in this version:

• Web UI for browser-based interaction
• AI conversation mode powered by Claude
• just command runner for easy task automation
• uv for fast Python package management

Requirements

• Python 3.8+
• uv - Fast Python package manager
• just - Command runner
• Claude Code - Required for AI conversation mode (optional)
• Bun - Required for Voice Generator local setup (optional)
• Docker - For containerized deployment (optional)

Quick Start

git clone https://github.com/HelgeSverre/glados.git
cd glados

# Install dependencies and download models
just setup

# Start the web server
just serve

Then open http://localhost:8765 in your browser.

Commands

Command	Description
just setup	Install dependencies and download models
just install	Install Python dependencies with uv
just download-models	Download and extract models from Google Drive
just serve	Start web server
just say "text"	Say something in GLaDOS's voice
just speak "text"	Have GLaDOS respond to you (AI mode with Claude)
just clean	Clean up generated audio files
just docker-build	Build Docker image
just docker-up	Start Docker container
just docker-up-detached	Start Docker container in background
just docker-down	Stop Docker container
just docker-logs	View Docker logs

Voice Generator Commands

Command	Description
just voice-setup	Install Bun dependencies for voice generator
just voice-start	Start both web server and TTS worker
just voice-serve	Start voice generator web server only
just voice-worker	Start TTS worker only
just voice-dev	Start web server in development mode (hot reload)
just voice-clean	Clean generated audio files and database
just voice-docker-build	Build voice generator Docker image
just voice-docker-up	Start voice generator Docker container
just voice-docker-up-detached	Start voice generator container in background
just voice-docker-down	Stop voice generator Docker container
just voice-docker-logs	View voice generator Docker logs

Visualizer Commands

Command	Description
just viz	Run aperture eye visualizer (watches for new audio)
just viz-file <file>	Run aperture eye visualizer with specific audio file

Docker

Run GLaDOS in a Docker container:

# Build and start
just docker-build
just docker-up

# Or in one command with docker compose
docker compose up --build

Then open http://localhost:8765 in your browser.

To run in background:

just docker-up-detached
just docker-logs  # view logs
just docker-down  # stop

Note: AI conversation mode requires Claude Code CLI and is not available in Docker. The Docker version supports TTS-only mode through the web interface.

Voice Generator

The Voice Generator is a standalone web application for batch generating GLaDOS voice lines. It features an Aperture Science-themed UI where you can queue up text, and a background worker processes them into audio files.

Features:

• Web UI for submitting text to synthesize
• Background TTS worker processes jobs from a queue
• SQLite database for job tracking
• Generated audio files can be played in-browser or downloaded
• Fully containerized with Docker support

Local Setup

Requires Bun for the web server:

# Install dependencies
just voice-setup

# Start both web server and TTS worker
just voice-start

Then open http://localhost:3000 in your browser.

Docker Setup

Run the voice generator in a Docker container with no host dependencies:

# Build and start
just voice-docker-build
just voice-docker-up

# Or with docker compose directly
cd voice-generator && docker compose up --build

To run in background:

just voice-docker-up-detached
just voice-docker-logs  # view logs
just voice-docker-down  # stop

The Docker container includes both the Bun web server and Python TTS worker. Generated audio and the SQLite database are persisted via volume mounts.

Visualizer

An audio-reactive aperture eye visualizer that responds to GLaDOS's voice. The iris contracts and expands based on audio amplitude, creating an immersive visual experience.

Keyboard Controls:

• SPACE - Replay audio
• ESC - Quit
• Drag-and-drop WAV files onto the window to play them

About the TTS Models

The neural network TTS models are from the original glados-tts project. See that repo for details on training data and model architecture.

Manual Installation

If you prefer manual setup:

1. Download the model files from Google Drive and unzip into the repo folder
2. Install dependencies: uv pip install -r requirements.txt
3. Run the server: uv run python web/server.py