# Dear Diary AI ## About - 文字をタイプすると自動生成された曲が流れる - ビジュアルも変化する - タイピングの画面と曲をシェアできる - オープンソース [Dear Diary AI](https://deardiary.ai/) > We created this for a ML & Music hackathon in our free time in August 2020. [Dear Diary | Devpost](https://devpost.com/software/dear-diary-ezrmgt) > We want you to journal and to feel better. [Studies](https://pubmed.ncbi.nlm.nih.gov/23790815/) [show](https://pubmed.ncbi.nlm.nih.gov/16942980/) that journaling increases mindfulness and alleviates stress, anxiety, and depression. > > We also want you to create music. Like journaling, the mental health benefits of creating music are [well-documented](https://www.sciencedirect.com/science/article/abs/pii/S0272735809000026). And music is awesome. ## More  ## How it works > **Sentiment Analysis** As you type each word, we perform sentiment analysis using the popular natural language processing library VADER.js. This returns a prediction score from -1.0 to +1.0, that we interpret as a sentiment along the spectrum of negative to neutral to positive. > **Music Creation using Magenta** We use Google’s [[Magenta]] MusicVAE pre-trained model to interpolate between two pre-composed melodies—one happy, one sad—created by our teammate Devin Lane who has a background in music composition and education. As you type, our app creates new music that lives in the latent space of the emotion spectrum between happy, neutral, and sad based on our sentiment analysis. (The rest is omitted)