hiyorimi/speech2textbot

Speech2Text bot (Telethon, YandexSpeechKit API)

Speech2Text telegram bot (via telepot) v.2.0

This is light demo of Telethon, Yandex Speech Kit and telegram capabilities to improve your life a bit. You can run this bot and recognize annoying audio messages to text if you dislike them or simply can't listen right away.

Demo

You can try @sp2txt_bot yourself. Just record a voice message, or forward one and wait couple of seconds.

Installation

1. Get your Yandex.SpeechKit API key at Yandex Developer Center. Paste it inside
2. Make sure Bot Father approves of your intentation to create another bot. If so, get a token.
3. Install dependencies required for audio processing.

$ sudo apt-get update && apt-get install ffmpeg 
# or, for Mac
$ brew install ffmpeg

Installation tested on Debian Wheezy and Ubuntu 16.04. If you have troubles with opus-tools, check out notes below.

4. Install dependancies and run bot with:

$ git clone https://github.com/Hiyorimi/speech2textbot.git
$ cd speech2textbot
$ cp env_example .env
$ pipenv shell
(speech2textbot)$ pipenv install

You can also paste key directly in code, it's a bit dirty. In that case, provide a token when launching a bot

$ sed -i "s/SPEECH_KIT_API\ \=\ ''/SPEECH_KIT_API\ \=\ 'your-yandex-speech-kit-api-here'/g" speech2textbot.py

On Mac you should add '' after -i:

$ sed -i '' "s/\=\ ''/\=\ 'your-yandex-speech-kit-api-here'/g" speech2textbot.py

Running

(speech2textbot)$ python speech2textbot.py

Docker

$ docker build -t speech2txtbot_docker .
$ docker run -d --name=speech2txtbot speech2txtbot_docker

Docker debug

docker run -d --name=speech2txtbot speech2txtbot_docker /bin/sh -c "while true; do sleep 2; df -h; done"
docker exec -it speech2txtbot /bin/sh

Final

I would be grateful for any ideas for improvement, feel free to fork, copy and modify.

Analytics

You can register your BOTAN_TOKEN at @Botaniobot to track bot users. Don't forget to include token into your .env file.

OPUS installation

If you can't build audiotools and fail to install opus-tools and opusfile from packages, you might want to build OPUS from source.

Since there is no installation guide, some tips. First of all, install system libs:

$ apt-get install vorbis-tools libogg-dev  libmpg123 libmp3lame-dev libmp3lame0 ffmpeg

1. Get opus, opus-tools and opusfile archives, unpack them with tar -xzf [file]
2. cd into directory and run

$ ./configure
$ ./make
$ ./make install

You might need to install additional dependencies.