Whisper CLI demos

Prepare environment

Install openai-whisper:

!pip show -qq openai-whisper || pip install --user git+https://github.com/openai/whisper.git

Note: this step duplicates setup performed in WhisperDemo.ipynb, so may not be needed.

Configure Bash command search path:

echo 'export PATH=~/.local/bin:$PATH' >> ~/.bash_profile && . ~/.bash_profile

Transcription examples

whisper --output_format txt --task transcribe --language en Winston_Churchill_-_Be_Ye_Men_of_Valour.wav
whisper --output_format txt --task transcribe --language en khosla.wav

Monitoring GPU usage during execution

In a separate terminal window, run:

watch -n 1 nvidia-smi

You'll see the model being uploaded to GPU ram followed by GPU core activity as it's executed.