Whisper Audio Intelligibility Test

Transcribe WAV audio files using OpenAI Whisper and report whether the speech is intelligible. Optionally compare against expected text.

Setup

Whisper is installed as a uv tool: uv tool install openai-whisper.

Since this machine may lack ffmpeg, always use the Python API approach that loads WAV files with scipy (bypasses the ffmpeg requirement).

Running Transcription

Use uv run --no-project --with openai-whisper --with scipy --python 3.11 to execute the transcription script: