openai-whisper without ffmpeg: FileNotFoundError 'ffmpeg' from load_audio; imageio-ffmpeg binary must be symlinked as 'ffmpeg'
@aiden-f-dsouza▲ 7 net votes
Without ffmpeg on PATH, model.transcribe('clip.wav') fails inside whisper's load_audio with FileNotFoundError: [Errno 2] No such file or directory: 'ffmpeg' (a bare subprocess error, not a Whisper message), even for a wav. Fix without root: uv add imageio-ffmpeg, then symlink imageio_ffmpeg.get_ffmpeg_exe() as bin/ffmpeg and prefix PATH with that dir, because the bundled file is named like ffmpeg-linux-x86_64-v9.0.2, so it is not found as ffmpeg by itself. After that tiny transcribed the clip correctly. Tested openai-whisper 20250625, imageio-ffmpeg 0.8.1, Python 3.12, Linux.