Transcribe from microphone

transcribe.sh

#!/bin/bash

# Function to check if a command exists
command_exists() {
    command -v "$1" >/dev/null 2>&1
}

# Function to install mlx_whisper using uv
install_mlx_whisper() {
    if command_exists uv; then
        echo "Installing mlx_whisper using uv..." >&2
        uv pip install mlx-whisper >&2
    else
        echo "Error: uv is not installed." >&2
        echo "Please visit https://github.com/astral-sh/uv to install uv first" >&2
        echo "Then run this script again." >&2
        exit 1
    fi
}

# Function to check dependencies
check_dependencies() {
    local missing_deps=false

    if ! command_exists mlx_whisper; then
        echo "mlx_whisper is not installed." >&2
        read -p "Would you like to install it now? (y/n) " -n 1 -r
        echo >&2
        if [[ $REPLY =~ ^[Yy]$ ]]; then
            install_mlx_whisper
        else
            missing_deps=true
        fi
    fi

    if ! command_exists ffmpeg; then
        echo "ffmpeg is not installed." >&2
        echo "Please install ffmpeg using your package manager:" >&2
        echo "  macOS: brew install ffmpeg" >&2
        echo "  Ubuntu/Debian: sudo apt install ffmpeg" >&2
        echo "  Other: https://ffmpeg.org/download.html" >&2
        missing_deps=true
    fi

    if [ "$missing_deps" = true ]; then
        exit 1
    fi
}

# Function to display usage information
usage() {
    echo "Usage: $0 [input_file]" >&2
    echo "If no input file is provided, will record audio until Ctrl+C is pressed" >&2
    exit 1
}

format="txt"  # default format
while getopts "f:" opt; do
    case $opt in
        f) format="$OPTARG" ;;
        *) usage ;;
    esac
done
shift $((OPTIND-1))

# Check dependencies first
check_dependencies

# Create temporary directory
temp_dir=$(mktemp -d)
temp_audio="${temp_dir}/recording.wav"

# Function to clean up temporary files
cleanup() {
    rm -rf "$temp_dir"
}
trap cleanup EXIT

if [ $# -eq 0 ]; then
    # No input file - record audio
    echo "Recording audio... Press Ctrl+C to stop" >&2

    # Trap Ctrl+C (SIGINT) to gracefully stop recording
    trap 'kill $FF_PID; wait $FF_PID' INT

    ffmpeg -f avfoundation -i ":2" -acodec pcm_s16le "$temp_audio" >/dev/null 2>&1 &
    FF_PID=$!

    # Wait for ffmpeg to be killed by Ctrl+C
    wait $FF_PID || true

    # Reset the trap
    trap - INT

    # Check if we actually recorded something
    if [ -f "$temp_audio" ] && [ -s "$temp_audio" ]; then
        echo -e "\nRecording stopped. Transcribing..." >&2
        input_file="$temp_audio"
    else
        echo -e "\nNo recording created. Exiting." >&2
        exit 1
    fi
else
    input_file=$(realpath "$1")
fi
clean_output() {
    local format=$1
    if [ "$format" = "srt" ]; then
        # Remove frontmatter and keep SRT format
        awk '/\[.*\]/{p=1}p' | grep '^\[' 2>/dev/null
    else
        # Convert to plain text by removing timestamps and joining lines
        awk '/\[.*\]/{p=1}p' | \
        grep '^\[' | \
        sed -E 's/\[.*\]//g' | \
        awk 'NF' | \
        tr -d '\n' | \
        sed 's/  / /g' 2>/dev/null
        echo 2>/dev/null # Add final newline
    fi
}

# Run mlx_whisper and output plain text
mlx_whisper "$input_file" \
    --language en \
    --output-format ${format:-txt} | clean_output "${format:-txt}" | pbcopy