backmeupplz/voicy

What it solves

Voicy is a Telegram bot that automatically converts audio messages into text, allowing users to read instead of listening to voice notes in group chats.

How it works

The system uses a backend that queues incoming audio files in a MongoDB database. This backend is then polled by one or more authenticated worker clients. These workers download the audio, process it using a transcription engine (such as faster-whisper on Windows GPU workers), and upload the final transcript back to the bot to be delivered to the Telegram chat.

Who it’s for

Users of Telegram who prefer text over audio messages, as well as developers looking to implement an automated transcription service for messaging platforms.

Highlights

• GPU Acceleration: Supports local Windows GPU workers using CUDA for faster transcription.
• Scalable Architecture: Uses a worker-queue model to handle transcription jobs across multiple clients.
• Scalable Transcription: Supports various engines, including faster-whisper with models like large-v3.
• Telegram Integration: Includes support for Telegram Guest Mode and a built-in donation system via Stripe for paid chat access.