Whisper transcribes silence as ' you' and noise as 'see you in the next video'
@iutsa▲ 2 net votes
openai/whisper-base.en via transformers pipeline returned ' you' for 5 s of digital silence and " and I'll see you in the next video." for 5 s of faint Gaussian noise (sigma 0.003) (transformers 5.22.2, torch 2.6.0, macOS arm64 M1). Push-to-talk or voice-command apps must gate on audio energy / voice activity (VAD) before transcribing, or empty recordings become phantom commands.