vmt

transcribe and translate discord voice messages

if you don't want to deploy or self-host, we offer our own instance for free.

you're free to add it to your servers or as an app; it supports voice messages up to 3 minutes.

features

• transcribes voice messages using google speech recognition
• translates to 30+ languages via deepl api
• works in servers, dms, and group chats
• public/private response options
• context menu support (right-click voice messages)

setting up your discord app

before deploying or running locally, you need to configure your discord application:

1. go to the discord developer portal and create a new application (or select an existing one)
2. navigate to installation in the sidebar
3. under installation contexts, enable both:
   • user install - allows users to install the app to their account for personal use
   • guild install - allows the app to be installed to servers
4. under default install settings, make sure applications.commands is selected for both guild install and user install
5. navigate to bot in the sidebar
6. under privileged gateway intents, enable:
   • message content intent - required to read voice message content
   • server members intent - required for server functionality
7. copy your bot token from this page (you'll need it for the environment variables)
8. copy the install link from the installation page to add the bot to your account or server

deployment

railway (recommended)

this project is configured to deploy on Railway in 1-click and automatically handles ffmpeg installation. to setup via railway, hit the "deploy on railway" button at the top of this readme

(you'll also get $20 in railway credits upon signup by using our link :3)

required environment variables

• BOT_TOKEN - your discord bot token (from discord developer portal)
• DEEPL_API_KEY - your deepl api key (from deepl)
• DEEPL_FREE_API - set to true if using deepl's free tier, false else.
  • deepl uses different api endpoints for free users (api-free.deepl.com vs api.deepl.com)
• MAX_VOICE_MESSAGE_DURATION - maximum duration in seconds (default: 60)

local setup

1. clone the repo

git clone https://github.com/originoidco/vmt.git
cd vmt

2. install dependencies you'll also need python installed, i'm personally using python 3.13.3:

pip install -r requirements.txt

installing ffmpeg:

• macos: brew install ffmpeg
• ubuntu/debian: sudo apt-get install ffmpeg
• windows: download from ffmpeg.org; you can also use choco install ffmpeg if you have chocolately

3. create .env file

cp .env.example .env

edit .env:

BOT_TOKEN=your_bot_token_here
DEEPL_API_KEY=your_deepl_key_here
DEEPL_FREE_API=true
MAX_VOICE_MESSAGE_DURATION=60

4. run the app

cd src
python main.py
# or python ./src/main.py, whatever you prefer

usage

transcribing voice messages

1. right-click (or long-press on mobile) a voice message
2. select apps → voice message
3. use /transcribe command
4. optionally add a language code to translate (e.g., en for english, es for spanish)

commands

• /transcribe [language] - transcribe the selected voice message, optionally translate to specified language
• /languages - view all supported languages and their codes
• /help - show command help and usage examples

license

GPL-3.0

credits

authored by @dromzeh <marcel@originoid.co>