Using pre-recorded speech instead of TTS