Pick the TTS voice a reply is spoken in — cards or a combobox, language and accent chips, neutral timbre tags, and a play button that auditions without selecting, with exactly one sample sounding at a time.