DeepSpeech inference Telegram bot
A simple telegram bot that will convert a voice note to text using a DeepSpeech model. Created to quickly test my DeepSpeech Catalan model.
Configuration
The following environment variables allow configuring the bot:
| Variable | Description | Default |
|---|---|---|
TELEGRAM_TOKEN |
The unique token of your bot | None |
MODEL_PATH |
The path to the DeepSpeech model | ./model/model.pbmm |
SCORER_PATH |
The path to the KenLM scorer | ./model/kenlm.scorer |
Usage
Install the requirements
$ pip install -r requirements.txt
and start the bot
$ python deepspeech-bot.py
Usage with Docker
Build the docker image
$ docker build -f Dockerfile -t deepspeech-bot .
Run the docker image. You can specify a volume and environment variables (please refer to the docker run documentation):
$ docker run -v /path/to/my/model:/app/model -e TELEGRAM_TOKEN deepspeech-bot