Transcribe mic audio using Whisper

generate.sh

#!/bin/bash

# record some raw audio
sox -d rec.wav

# resample to 16kHz
ffmpeg -y -i ./rec.wav -ar 16000 -ac 1 -c:a pcm_s16le ./rec16.wav > /dev/null 2>&1

# run Whisper
echo "Processing ..."
./main -m models/ggml-base.en.bin rec16.wav -owts > /dev/null 2>&1

# generate Karaoke video
source rec16.wav.wts > /dev/null 2>&1

# play the video
ffplay -loglevel 0 -autoexit ./rec16.wav.mp4