Module audio_processor 

Audio preprocessing for Whisper ASR.

Load audio files → decode → resample to 16kHz mono → f32 PCM samples. Supports WAV natively; M4A/MP3/FLAC/OGG via ffmpeg auto-conversion.

Constants

CHUNK_SAMPLES

Whisper processes 30-second chunks. At 16kHz → 480,000 samples.

Functions

chunk_pcm

Split PCM samples into 30-second chunks for Whisper processing.

load_audio

Load audio file and return 16kHz mono f32 PCM samples.

load_audio_at_rate

Load audio file and return mono f32 PCM samples at a configurable sample rate.

load_audio_bytes

Load audio from raw bytes. Tries WAV first; if that fails and bytes look non-WAV, tries ffmpeg.