日本語

Discord Text-to-Speech Bot

An open-source Discord TTS bot written in Rust. It reads aloud Japanese text chat messages using VOICEVOX ENGINE, providing a lightweight self-hosted alternative to existing TTS bots.

Features

• Natural Japanese speech synthesis powered by VOICEVOX ENGINE
• Per-server dictionary for custom word pronunciations
• Slash command interface (/v, /s, /dict, /help)
• Auto-join voice channels on message
• Hourly time announcements

Prerequisites

Hardware

Speech synthesis is CPU/GPU intensive. A reasonably powerful machine with at least 2 GB of RAM is recommended. Latency will be higher on slower hardware, especially right after startup while the VOICEVOX model initializes.

Software

Dependency	Purpose	Notes
libopus-dev	Opus audio encoding	Required at build time
ffmpeg	Audio format conversion	Required at runtime
Redis	Persistent storage (dictionaries, settings)	v7+ recommended
VOICEVOX ENGINE	Japanese speech synthesis	CPU or GPU version

Discord Bot Setup

1. Open the Discord Developer Portal and create a new application.
2. Note the Application ID (Client ID).
3. Go to the Bot page, click Add Bot, and enable Message Content Intent.
4. Click Reset Token to generate a token — save it securely.
5. Invite the bot to your server with the following URL (replace CLIENT_ID):

https://discord.com/api/oauth2/authorize?client_id=CLIENT_ID&permissions=3146752&scope=bot%20applications.commands

Required permissions: View Channels, Connect, Speak.

Installation

Docker Compose (recommended)

cd deployment

# Create .env from template and fill in your values
cp config/.env.template .env
vi .env

# Start all services
docker compose up -d

# View logs
docker compose logs -f

# Stop
docker compose down

Native Build

# Install system dependencies (Debian/Ubuntu)
sudo apt-get update
sudo apt-get install -y build-essential pkg-config libopus-dev ffmpeg

# Build
cargo build --release

# The binary is at target/release/bot
# Make sure Redis and VOICEVOX ENGINE are running, then:
export DISCORD_CLIENT_ID="your_client_id"
export DISCORD_BOT_TOKEN="your_bot_token"
export REDIS_URL="redis://localhost:6379"
export VOICEVOX_API_BASE="http://localhost:50021"

./target/release/bot

NixOS

# Enter development shell
nix develop

# Or build the bot directly
nix build
./result/bin/bot

systemd

A systemd unit file is provided at deployment/voicebot.service.

# Create a dedicated user
sudo useradd --system --user-group voicebot

# Install the binary
sudo cp target/release/bot /usr/local/bin/bot

# Set up environment file
sudo mkdir -p /etc/voicebot
sudo vi /etc/voicebot/env   # Add DISCORD_CLIENT_ID, DISCORD_BOT_TOKEN, etc.

# Install and start the service
sudo cp deployment/voicebot.service /etc/systemd/system/
sudo systemctl daemon-reload
sudo systemctl enable --now voicebot

Configuration

The bot reads configuration from environment variables. If they are not set, it falls back to a YAML config file (default: /etc/bot.yaml, override with BOT_CONFIG).

Variable	Required	Default	Description
DISCORD_CLIENT_ID	Yes	—	Discord application ID
DISCORD_BOT_TOKEN	Yes	—	Discord bot token
REDIS_URL	Yes	—	Redis connection URL
VOICEVOX_API_BASE	No	http://voicevox:50021	VOICEVOX ENGINE endpoint
RUST_LOG	No	info	Log level filter (env_logger)
SENTRY_DSN	No	—	Sentry error tracking DSN
BOT_CONFIG	No	/etc/bot.yaml	Path to YAML config file (fallback)

Usage

Command	Description
/v	Join / leave the voice channel (toggle)
/s	Skip the current message being read
/dict add <word> <reading>	Add a word to the server dictionary
/dict remove <word>	Remove a word from the dictionary
/dict list	Show all dictionary entries
/autojoin	Toggle auto-join for the current channel
/time	Toggle hourly time announcements
/help	Show help information

Architecture

Discord message
    │
    ▼
Text processing ── URL/emoji removal, dictionary replacement, 60-char limit
    │                (crates/bot)
    ▼
VOICEVOX API ───── /audio_query → /synthesis
    │                (crates/bot-speech)
    ▼
EncodedAudio
    │
    ▼
FFmpeg ─────────── Convert to PCM (48kHz, mono, 16-bit signed LE)
    │                (crates/bot-audio)
    ▼
DecodedAudio
    │
    ▼
Songbird queue ─── Encode & transmit to Discord voice channel
                     (crates/bot-call)

Project Structure

crates/
├── bot/          # Main bot binary — event handling, commands, message processing
├── bot-audio/    # Audio encoding and FFmpeg integration
├── bot-call/     # Voice connection management
├── bot-config/   # Configuration loading (env vars / YAML)
├── bot-db/       # Redis-backed persistence (dictionaries, settings)
└── bot-speech/   # VOICEVOX API client and speech synthesis

License

MIT — see LICENSE for details.