Small script to transcribe voice messages

whisper

#!/bin/sh

# Install whisper.cpp from here: https://github.com/ggerganov/whisper.cpp
# You may also need ffmpeg for conversion

WHISPER_PATH=/Users/pavel/.local/share/whisper.cpp
MODEL="large"

usage()
{
  echo "Usage:"
  echo "    $0 [wav-file]"
  echo "    You can also use non-wav audio file, ffmpeg will be used for conversion then."
}

if [ "$#" -ne 1 ]; then
  echo "Error: wrong arguments count!"
  usage
  exit 1
fi

inputfile=$1

extension="${inputfile##*.}"
if [ "$extension" != "wav" ]; then
  echo "Warining: Not a wav-file! Trying to convert with ffmpeg." >& 2
  tempfile="$(mktemp).wav"
  ffmpeg -i $inputfile -ac 1 -ar 16000 $tempfile
  if [ "$?" -ne 0 ]; then
    echo "Error: failed to convert file to wav."
    usage
    exit 1
  fi
  inputfile=$tempfile
fi

$WHISPER_PATH/whisper -m $WHISPER_PATH/models/ggml-$MODEL.bin -f $inputfile -pc -l ru -nt -su 2> /dev/null
if [[ -n "${tempfile+x}" ]]; then
  rm $tempfile
fi