Beyond Text: Discovering Voice and Image Input in ChatGPT for Extra Use Cases