Fact · Transcription
Whisper's load_audio function decodes an audio file by starting the ffmpeg command-line tool as a subprocess, and it needs ffmpeg in the PATH to do so.
Checked against its source on October 17, 2026
In the source's own words
“This launches a subprocess to decode audio while down-mixing ... Requires the ffmpeg CLI in PATH.”
whisper/audio.py: the audio constants and log-Mel code in the openai/whisper repository, OpenAI, GitHub, openai/whisper repository (main branch), 2022, load_audio function in whisper/audio.py.
Open the original- What it measures
- The function passes the file to ffmpeg with the input option, then asks ffmpeg for mono 16-bit PCM audio at the model's sample rate. The code does not list the accepted input formats, so the formats are whatever ffmpeg can decode.
- What it does not show
- This describes the open-source Whisper code. Apps that wrap Whisper may load audio in their own way, so check the app's documentation for its formats.