When in new prompt, directly start recording of voice (according presets) and providing transcribed text afterwards into prompt.