Whisper does audio to text with translation, so you'd have to extract the audio channel first.