telegram-voice-to-text

@audiotranscript_bot

Feel free to write me @emilkayumov

Introduction

A simple telegram bot to transcript voice messages.

It's my pet project which is main goal not to be something big. I'm going to learn some basic stuff: telegram bot building, docker, databases (like sqlite if I imagine how to use it), clouds, tests, ci, moninorings, logging, etc.

You can deploy your own installation of this bot. It requires only bot token and some computation power (I test it on my Air m1 and it works without any problem with large model of Whisper).

I'm using now:

• Faster Whisper transcription with CTranslate2
• Python Telegram Bot

How to run

After creating a bot with @BotFather fill config.yaml with token. Also you can use allowlist_usernames if you don't want anybody to use your bot.

With virtualenv

You can create a virtual environment and setup it with:

make init

And run it (it could be long first time because of downloading a model):

make run

With Docker

You can build an image:

make docker-build

And start it:

make docker-run

Or use daemon mode:

make docker-run-daemon

Plans

1. Init
2. Run a bot on cloud
3. A docker file to easily run on server
4. Setup monitorings to see if the bot is crashed
5. Make the bot stable
6. Make test environment
7. Setup logging (no user messages only using stats)
8. Add some new functionality like summarizing messages
9. ...