Transcribe recordings with the C++ port of OpenAI Whisper

whisper.cpp.sh

# Clone the project
git clone https://github.com/ggerganov/whisper.cpp && cd whisper.cpp

# Download at least one model, which may be "tiny", "base", "small", "medium", 
# "large", etc. Usually "medium" is sufficient for recordings in Chinese.
bash ./models/download-ggml-model.sh medium

# Compile the program
make

# Prepare the input audio file, since the current version runs only with 16-bit
# WAV files. Dependent on ffmpeg.
ffmpeg -i input.mp3 -ar 16000 -ac 1 -c:a pcm_s16le test.wav

# Start transcription with the "medium" model (-m, or with another model of
# choice) and use Chinese (-l, for the full list of language codes see 
# https://github.com/openai/whisper/blob/main/whisper/tokenizer.py#L10).
./main -l zh -m models/ggml-medium.bin -f test.wav