A Telegram bot for transcribing audios

I hate chat audios… but I also love making them.
That’s a hell of a contradiction, isn’t it? 😅

Voice messages are convenient when you’re sending them… and kind of a pain when you’re receiving them. You have to stop, listen, rewind, find a quiet moment—it’s not always ideal.

Most chat apps try to fix this with transcription, but it’s usually locked behind a paywall, half-baked, or just not available everywhere.

Scriberr

Lately I’ve been betting quite hard on self-hosting and home hardware. Among the many services I run, there’s Scriberr, which is honestly a gem.

It lets you generate transcripts from pretty much anything: audio files, videos, YouTube links, you name it, and you can choose which model to run depending on your hardware.

And well… I have an RTX 4070 sitting in my PC doing nothing most of the time.

On top of that, Scriberr exposes a clean API. At that point, it was a nobrainer:
“Why not build something to deal with all these voice messages?”

Bots to the rescue 🤖

I’ve always loved bots. They’re such a simple but powerful way to extend a platform.

Telegram bots in particular are a different breed. I’ve been building them since 2016, and they’ve always felt surprisingly capable (I still remember playing mini-games on them).

So, this time, I built bot that transcribes audio.
You send it a voice message or a file, it uploads everything to Scriberr, and sends back the transcription right in the same chat.

The nice part is that Telegram is super open. You can forward audio from basically any other app as a file, drop it into the bot, and get your transcript back.

So in a way, this bot becomes a quick gateway to your own Scriberr instance: no friction, no extra steps, just forward and read.

It’s called ScriberrTG and this is the repo, it’s dockerized and published on ghcr, so the installation is as easy as possible as long as you’re a bit familiar with self-hosting.

Next time I’ll show you my home setup and give you some ideas if you plan starting one 😶‍🌫️