Simple curl script to play aloud what you type, useful if your voice is suddenly broken.

Features

• Hacked in 30 minutes
• Caching: if asking speech for something that's been spoken before, will play the previous file

Dependencies

• curl
• openai cli, and an API key setup
• jq
• mplayer
• optional: notify-send

Usage

• chmod +x speech.sh
• Optional: Put your OpenaiAPI key in a file called API_KEY in the same dir as this one or supply it as an argument.
• ./speech.sh --voice "alloy" --text "This is a test" --speed 1.1 --api_key ****...***
• If you want to add an i3wm binding: bindsym $mod+p exec /YOUR/PATH/speech_launcher.sh, mode "default" with speech_launcher.sh containing this:

#!/usr/bin/zsh
OUT=$(i3-input)
TEXT=$(echo $OUT | tail -n 1 | cut -d'=' -f 2 | cut -c2-)
API_KEY=$(cat /YOUR/PATH/speech.sh/API_KEY)
/YOUR/PATH/speech.sh/speech.sh --text "$TEXT" --api_key $API_KEY

Why ?

Ospeak was too slow because of the python overhead

Related project

• quick_whisper_type.py: Minimalist python script to turn your voice into typed text (works anywhere)