Whisper

This projects provides a helper script to transcribe audio/video files using the OG Whisper model from OpenAI

Installation

Create virtual environment and install dependencies

python -m venv venv
source venv/bin/activate
pip install -r requirements_lock.txt

Usage

Transcribe

Transcribes a file by automatically detecting the language

./transcribe.sh audio_or_video.mp4

Specifying the Audio Language

Use the -l or --language flag to specify the language code (e.g., ja for Japanese).

./transcribe.sh japanese.mp3 --language ja

Limiting Words Per Line

Use the -m or --max-words flag to limit the model output

./transcribe.sh audio_or_video.mp4 --language en --max-words 7

Note

The word limited output will be available only for timestamped files like .srt, .vtt, etc.

Translation Mode

Use the -t or --translate flag to transcribe and translate into English

./transcribe.sh korean.mp3 --language ko --translate

Debug

Use the --debug flag to print the exact shell command that will be executed without running it

./transcribe.sh file.mp4 --debug

Options Reference

Flag	Short	Description	Default
--language	-l	Specify the language code (e.g., en, fr)	Auto-detect
--max_words	-m	Sets the maximum number of words per line	Auto
--translate	-t	Instruct Whisper to translate the audio into English	Off
--debug	-d	Prints the underlying command without executing it	Off