We developed a perfume recommendation system that understands user queries. It accepts various types of text input, which could be the users' favorite notes and accords, the desired occasions to wear the perfumes, or even a piece of poem about the mood. The recommender returns a list of perfumes, with vibrant graphs displaying the basic properties of the perfumes, and generated text explanations for the recommendations. Our recommender is the first that deciphers the needs of the users from plain text, and justifies its solutions with data visualizations and brief descriptions. This project saves time and energy for the customers by lifting the burden of searching for a perfume on the internet, which could be a swamp of pointless comments and debates. New users are warmly welcomed to perfumes by the user-friendly app interface and easy to understand demonstrations, even without access to samples. Just enter a sentence, we make scents and it makes sense.