This is a simple Telegram bot which takes voice clips and responds with the transcription. Meant to be my personal pensieve-bot.
The ASR model exploration can be found in this STT-exploratory gist and in particular this implementation is based on OpenAI's whisper model
The bot integration uses pyTelegramBotAPI
NOTE: It is recommended to create a virtual env with python 3.8+
pip install -r requirements.txt
First create a .env
file with the following values:
VALID_SENDER_IDS="sender_id_1,sender_id_2,..." # telegram user IDs to which the bot replies
BOT_TOKEN="your-bot-token" # Telgram bot token (from BotFather)
The run with:
# Run the bot with the 'small' whisper model
python -m src.run -m small -v INFO