Microphone voice transcription and translation with OpenAI Whisper

whisper.sh

#!/bin/bash

# ====================================
# author: https://github.com/maledorak
# email: [email protected]
# version: 0.1
# description: Script for recording voice from microphone and sending to Whisper OpenAI
#
# how to use:
#   1. run it - this will start recording sound: ./whisper.sh -c transcribe -l 30
#   2. stop recording by pressing Ctrl+C or rerun it again with the same arguments to 'shorten' the recording
#
# tips:
#   The best way to use it is to bind it to a key combination in your window manager
#   and then just press it to start recording and stop recording with the same key combination
# ====================================

# Fail the script
function fail { exit 1; }

# Supported commands
TRANSCRIBE='transcribe'
TRANSLATE='translate'
COMMANDS_ARRAY=($TRANSCRIBE $TRANSLATE)
COMMANDS_PIPED=$(IFS="|" ; echo "${COMMANDS_ARRAY[*]}")

function show_help_and_exit
{
    printf "\nUsage: %s -c [${COMMANDS_PIPED}] -s [30] \n\n" "$(basename "$0")"
    printf "Script for recording voice from microphone and sending to Whisper OpenAI \n\n"
    printf "Options:\n"
    printf "  %-20s  - %s\n"  "-h|--help"        "Show this help message"
    printf "  %-20s  - %s\n"  "-c"               "Whisper command to run [${COMMANDS_PIPED}}]"
    printf "  %-20s  - %s\n"  "-l"               "Record length in seconds (default: 30)"
    printf "  %-20s  - %s\n"  "-k"               "Kill all whisper processes"
    printf "\n"
    fail
}

#====== Check if following commands are installed ====
# Check if ffmpeg is installed
if ! command -v ffmpeg &> /dev/null
then
    echo "ffmpeg could not be found"
    fail
fi

# Check if jq is installed
if ! command -v jq &> /dev/null
then
    echo "jq could not be found"
    fail
fi

# Check if xclip is installed
if ! command -v xclip &> /dev/null
then
    echo "xclip could not be found"
    fail
fi

# Check if openai is installed
if ! command -v openai &> /dev/null
then
    echo "openai could not be found"
    fail
fi

# Check if notify-send is installed
if ! command -v notify-send &> /dev/null
then
    echo "notify-send could not be found"
    fail
fi
#=======================================

# ----------------------------------------------------------------------------------------------------------------------

COMMAND=
LENGTH=30

# Parse arguments
while getopts "hkc:l:" opt; do
    case "${opt}" in
        h)
            show_help_and_exit
            ;;
        \?)
            error "Unknown option: $1"
            show_help_and_exit
            ;;
        c)
            COMMAND=${OPTARG}
            ;;
        l)
            LENGTH=${OPTARG}
            ;;
        k)
            notify-send "Whisper" "Killing all processes" &
            pkill -f whisper
            exit 0
            ;;
    esac
done

# Validate command argument
if [[ -z $COMMAND ]]; then
    show_help_and_exit
fi

# Validate command argument, only transcribe or translate are supported
if [[ ! "$COMMAND" =~ ^($COMMANDS_PIPED)$ ]]; then
    echo "You can use only [$COMMANDS_PIPED] as a command [-c] argument, but '${COMMAND}' was used!"
    show_help_and_exit
fi

# Validate length argument
if [[ ! $LENGTH =~ ^[[:digit:]-]+$ ]]; then
	echo "Length [-l] argument must be a number, but '${LENGTH}' was used!"
    fail
fi

# Validate length argument
if [[ -z $OPENAI_API_KEY ]]; then
    echo "OPENAI_API_KEY environment variable is not set!"
    echo "Use the following link to set it up: https://platform.openai.com/account/api-keys"
    fail
fi

function openai_request {
    local cmd=$1
    local file_path=$2

    declare -A cmd_map=( ["$TRANSCRIBE"]="transcriptions" ["$TRANSLATE"]="translations" )
    local url_path=${cmd_map[$cmd]}

    # https://platform.openai.com/docs/api-reference/audio/create
    local response=$(curl https://api.openai.com/v1/audio/$url_path \
            -H "Authorization: Bearer $OPENAI_API_KEY" \
            -H "Content-Type: multipart/form-data" \
            -F file="@$file_path" \
            -F model="whisper-1")

    local text=$(echo $response | jq -r '.text')
    if [ "$text" == "null" ]; then
        echo "Whisper API error $response"
        return 1
    else
        echo $text
        return 0
    fi
}

function ffmpeg_record {
    local file_path=$1
    ffmpeg \
        -f alsa \
        -channels 2 \
        -sample_rate 44100 \
        -i hw:0 \
        -t $LENGTH \
        $file_path

    local code=$?
    if [ $code -eq 255 ]; then
        # ffmpeg returns 255 when Ctrl+C is pressed or SIGINT (-2) is received
        echo "FFmpeg has stopped with SIGINT"
        return 0
    elif [ $code -ne 0 ]; then
        echo "FFmpeg error $code"
        return $code
    else
        echo "FFmpeg has stopped"
        return 0
    fi
}

FILE_PREFIX=whisper_$COMMAND
IS_OTHER_INSTANCE_RUNNING=$(pgrep -fa $FILE_PREFIX | awk '{print $NF}')

if [ -z "$IS_OTHER_INSTANCE_RUNNING" ]; then
    echo "No process, FFmpeg recording"
    notify-send "Whisper" "$COMMAND is listening..." &

    FFMPEG_OUT="/tmp/${FILE_PREFIX}_$(date '+%Y%m%d%H%M%S').mp3"
    ffmpeg_record $FFMPEG_OUT
    ffmpeg_record_code=$?
    if [ $ffmpeg_record_code -eq 0 ]; then
        echo "FFmpeg has stopped, sending request to OpenAI"

        declare -A cmd_map=( ["$TRANSCRIBE"]="Transcribing" ["$TRANSLATE"]="Translating" )
        notify_string=${cmd_map[$COMMAND]}

        notify-send "Whisper" "$notify_string..." &
        text=$(openai_request $COMMAND $FFMPEG_OUT)
        openai_code=$?
        if [ $openai_code -ne 0 ]; then
            notify-send "Whisper" "OpenAI error..." &
            rm -rf $FFMPEG_OUT
            exit $openai_code
        else
            echo "OpenAI response: $text"
            echo -n $text | xclip -selection clipboard
            notify-send "Whisper" "'$text' copied to clipboard." &
            echo "Clean up file $FFMPEG_OUT"
            rm -rf $FFMPEG_OUT
            exit 0
        fi
    else
        echo "FFmpeg error $ffmpeg_record_code"
        notify-send "Whisper" "FFmpeg error $ffmpeg_record_code" &
        exit $ffmpeg_record_code
    fi
else
    echo "Process found, killing and trimming audio"

    pid=$(pgrep -fa $FILE_PREFIX | awk '{print $1}')
    kill -SIGINT $pid
    while $(kill -0 $pid 2>/dev/null); do
        echo 'Waiting for ffmpeg to finish...'
        sleep 0.01
    done
fi