speech2text

#!/data/data/com.termux/files/usr/bin/bash
DEEPSPEECH_VERSION=0.9.3
SPEECH_TIME=2

ARCH="$(uname -m)"
if [[ "$ARCH" == "aarch64" ]]; then
    ARCH="arm64"
elif [[ "$ARCH" == "armv7l" ]]; then
    ARCH="armv7"
fi

SOURCE_DIR="$(pwd)"

enterScriptDir() {
    SOURCE="${BASH_SOURCE[0]}"
    while [ -h "$SOURCE" ]; do # resolve $SOURCE until the file is no longer a symlink
        DIR="$( cd -P "$( dirname "$SOURCE" )" >/dev/null 2>&1 && pwd )"
        SOURCE="$(readlink "$SOURCE")"
        [[ $SOURCE != /* ]] && SOURCE="$DIR/$SOURCE" # if $SOURCE was a relative symlink, we need to resolve it relative to the path where the symlink file was located
    done
    DIR="$( cd -P "$( dirname "$SOURCE" )" >/dev/null 2>&1 && pwd )"
    cd "$DIR"
}

#Ensure we're in the script's directory
enterScriptDir

export LD_LIBRARY_PATH="$(pwd)/deepspeech-bin/"

#Setup
command -v speech2text > /dev/null || ln -s "$(pwd)/speech2text" "$HOME/../usr/bin/speech2text" 
ensure_requirements_installed() {
    command -v ffmpeg > /dev/null || pkg i -y ffmpeg
    #command -v mpv > /dev/null || pkg i -y mpv
    command -v wget > /dev/null || pkg i -y wget
    command -v termux-microphone-record > /dev/null || pkg i -y termux-api
}
download_deepspeech() {
    rm -rf ./deepspeech-bin
    mkdir ./deepspeech-bin
    cd ./deepspeech-bin
    wget "https://github.com/mozilla/DeepSpeech/releases/download/v${DEEPSPEECH_VERSION}/native_client.${ARCH}.cpu.android.tar.xz" -O ./native_client.tar.xz -q --show-progress
    tar xf ./native_client.tar.xz
    rm -f ./native_client.tar.xz
    cd ..
}

download_example_audio() {
    rm -rf ./audio
    wget "https://github.com/mozilla/DeepSpeech/releases/download/v${DEEPSPEECH_VERSION}/audio-${DEEPSPEECH_VERSION}.tar.gz" -O ./example-audio.tar.gz -q --show-progress
    tar xf ./example-audio.tar.gz
    rm -f ./example-audio.tar.gz
}

download_models() {
    rm -rf ./models
    wget "https://github.com/mozilla/DeepSpeech/releases/download/v${DEEPSPEECH_VERSION}/deepspeech-${DEEPSPEECH_VERSION}-models.tflite" -O "./deepspeech-${DEEPSPEECH_VERSION}-models.tflite" -q --show-progress && \
    wget "https://github.com/mozilla/DeepSpeech/releases/download/v${DEEPSPEECH_VERSION}/deepspeech-${DEEPSPEECH_VERSION}-models.scorer" -O "./deepspeech-${DEEPSPEECH_VERSION}-models.scorer" -q --show-progress
    if [[ "$?" != 0 ]]; then
        wget "https://github.com/mozilla/DeepSpeech/releases/download/v${DEEPSPEECH_VERSION}/deepspeech-${DEEPSPEECH_VERSION}-models.tar.gz" -O ./models.tar.gz -q --show-progress
        echo "Extracting ./models.tar.gz ..."
        tar xf ./models.tar.gz
        rm -f ./models.tar.gz
        mv ./deepspeech-*-models ./models
        rm -f ./._*
        rm -f ./models/._*
    fi
}
ensure_requirements_installed
#if [ ! -f ./deepspeech-bin/deepspeech ]; then
#    download_deepspeech
#fi

DETECTED_DEEPSPEECH_VERSION=$(LD_LIBRARY_PATH="$LD_LIBRARY_PATH" ./deepspeech-bin/deepspeech --version 2>&1 | grep DeepSpeech | cut -d "v" -f2 | cut -d "-" -f1 | sed -e 's/DeepSpeech //')

if [ "$DEEPSPEECH_VERSION" != "$DETECTED_DEEPSPEECH_VERSION" ]; then
    download_deepspeech
    download_example_audio
    download_models
fi

ls ./audio/*.wav &> /dev/null || download_example_audio

if [ ! -f ./models/output_graph.tflite ] && [ ! -f "./deepspeech-${DEEPSPEECH_VERSION}-models.tflite" ]; then
    download_models
fi

# Important part of the script

mic_listen() {
    rm -f ./tmp.mp4
    rm -f ./tmp.wav
    OUTPUT="$(termux-microphone-record -f ./tmp.mp4 -l $SPEECH_TIME -r 16000 -c 1 -e amr_nb)"
    if [[ "$OUTPUT" =~ "Recording error: null" ]]; then
        termux-notification --sound -t "Speech Recognition Error" -c "Can't access the mic. Make sure not other app is using the mic."
        exit
    fi
}
recognize() {
    termux-microphone-record -q > /dev/null
    ffmpeg -loglevel panic -i ./tmp.mp4 -f wav -bitexact -acodec pcm_s16le -ar 16000 -ac 1 ./tmp.wav
    rm -f ./tmp.mp4
    termux-media-player play ./tmp.wav > /dev/null & #play recording back before deepspeech

    if [[ "$DEEPSPEECH_VERSION" == "0.5.1" ]]; then
        ./deepspeech-bin/deepspeech --model ./models/output_graph.tflite --alphabet ./models/alphabet.txt --lm ./models/lm.binary --trie ./models/trie --audio ./tmp.wav 2>&1 | tail -1
    elif [ -f "./deepspeech-${DEEPSPEECH_VERSION}-models.tflite" ] ; then
        ./deepspeech-bin/deepspeech --model "./deepspeech-${DEEPSPEECH_VERSION}-models.tflite" --scorer "./deepspeech-${DEEPSPEECH_VERSION}-models.scorer" --audio ./tmp.wav 2>&1 | tail -1
    elif [ -f ./models/output_graph.tflite ] ; then
        ./deepspeech-bin/deepspeech --model ./models/output_graph.tflite --lm ./models/lm.binary --trie ./models/trie --audio ./tmp.wav 2>&1 | tail -1
    else
        echo "Error: Couldn't find the deepspeech models!"
    fi
    rm -f ./tmp.wav
}


mic_listen
sleep $SPEECH_TIME
recognize

cd "$SOURCE_DIR"