scrapyard-cogworks/ttstoy/dectalk-server/README.md

DECTalk TTS Server

A simple Node.js server that provides DECTalk text-to-speech with MiniMax-compatible API endpoints for RedBot TTSTOY cog.

Features

• Authentic DECTalk voice (Perfect Paul - same as Moonbase Alpha)
• MiniMax-compatible API endpoint (/v1/t2a_v2) for seamless integration with TTSTOY
• Simple GET endpoint (/say?text=<text>) for testing
• Automatic WAV to MP3 conversion using ffmpeg
• Health check endpoint (/health)

Requirements

• Node.js (v14 or higher)
• npm
• ffmpeg (for audio conversion)
• Linux system (DECTalk package doesn't support macOS)

Installing Dependencies

Ubuntu/Debian:

sudo apt install nodejs npm ffmpeg

Fedora:

sudo dnf install nodejs npm ffmpeg

Arch Linux:

sudo pacman -S nodejs npm ffmpeg

macOS:

brew install node ffmpeg
# macOS has built-in 'say' command, no need for espeak

Installation

cd dectalk-server
npm install

This will install:

• express - Web server framework
• dectalk - Authentic DECTalk voice synthesis (Moonbase Alpha)

The dectalk package includes the actual DECTalk binaries and will work on Linux systems.

Usage

Start the server

npm start

The server will run on port 3000 by default.

Custom port

PORT=8080 npm start

Custom output directory

OUTPUT_DIR=/path/to/output npm start

API Endpoints

Health Check

GET /health

Returns:

{"status": "ok"}

Engine Info

GET /engine

Returns:

{
  "type": "dectalk",
  "version": "1.0.0",
  "description": "DECTalk Text-to-Speech Engine"
}

Simple TTS (GET)

GET /say?text=Hello%20world

Returns: MP3 audio file

MiniMax-Compatible TTS (POST)

POST /v1/t2a_v2
Content-Type: application/json

{
  "text": "Hello world",
  "voice_setting": {
    "voice_id": "default",
    "speed": 1.0
  }
}

Returns:

{
  "base_resp": {
    "status_code": 0,
    "status_msg": "Success"
  },
  "data": {
    "audio": "<hex-encoded-mp3>",
    "audio_id": "<unique-id>"
  }
}

Integration with RedBot TTSTOY

1. Start the DECTalk server:

npm start

2. Configure TTSTOY to use DECTalk mode:

[p]ttstoy mode dectalk
[p]ttstoy dectalkurl http://127.0.0.1:3000

3. Test it:

[p]tts Hello, I am DECTalk

Notes

• Uses authentic DECTalk (Perfect Paul voice from Moonbase Alpha)
• The dectalk npm package includes the actual DECTalk binaries
• Audio files are saved to the output/ directory
• MP3 format: mono, 32kHz, 128kbps (matches TTSTOY expectations)
• Supports all 9 original DECTalk voices
• Supports DECTalk phoneme commands (e.g., [:phoneme on], [:rate 200])

DECTalk Commands

DECTalk supports special commands for controlling speech:

[:phoneme on] - Enable phoneme mode
[:rate 200] - Set speech rate (default 180)
[:dv ap 100] - Set voice parameters

Example:

curl "http://127.0.0.1:3001/say?text=%5B:rate%20150%5DHello%20world&voice=paul" -o test.mp3

You can also use these in Discord:

[p]tts [:rate 200] speaking very fast
[p]tts [:phoneme on] hehlow werld

DECTalk Voices

DECTalk includes 9 classic voices:

Voice	Name	Description
Paul	Perfect Paul	Default male voice (Moonbase Alpha, Stephen Hawking)
Betty	Beautiful Betty	Female voice
Harry	Huge Harry	Deep male voice
Frank	Frail Frank	Elderly male voice
Dennis	Doctor Dennis	Male voice
Kit	Kit the Kid	Child voice
Ursula	Uppity Ursula	Female voice
Rita	Rough Rita	Female voice
Wendy	Whispering Wendy	Soft female voice

Using Different Voices

In Discord:

[p]ttstoy voice Paul
[p]tts Hello, I am Perfect Paul
[p]ttstoy voice Betty
[p]tts Hello, I am Beautiful Betty

Direct API:

curl "http://127.0.0.1:3001/say?text=Hello&voice=paul" -o paul.mp3
curl "http://127.0.0.1:3001/say?text=Hello&voice=betty" -o betty.mp3

List available voices:

curl http://127.0.0.1:3001/voices

Customization

The server uses the default DECTalk voice (Perfect Paul). The dectalk package handles all the voice synthesis internally, so no additional configuration is needed for the authentic Moonbase Alpha sound.

Development

Run with auto-reload:

npm run dev

License

MIT