When Lyria 3 launched last month, Gemini could generate 30-second audio clips from prompts. With the new advanced model, the chatbot can now create tracks up to 3 minutes long and follow instructions ...