MoodTrack is a mobile application that encourages daily mood check-ins and recommends songs to users. The first iteration of the app used the Spotify API - users could login to their Spotify accounts via OAuth2.0, the app would randomly generate a list of songs in the background, filter these based on the mood slider values and present users with song recommendations. The sliders were based on the API’s Get Tracks’ Audio Features endpoint which included how suitable a track was for dancing, how fast and loud it was or its positiveness. These features or ‘moods’ are pre-determined by a Spotify algorithm and so not really translatable to human emotions. Still, there was definitely something behind it and the recommendations were mostly spot on…
Another feature I was able to build via the Spotify API was a playlist creator where users could take up to 100 of the latest song recommendations and add them to a Spotify playlist.
XRecorder_22092022_200753_1.mp4
The app also included a mood tracker which allowed users to select any seven day period and generate line charts based on their historic mood inputs.
All great, working features but I wasn’t totally convinced by the translation of the mood inputs from the Spotify audio features. I spoke to a dev friend about the project one day and he also asked “What if I don’t have Spotify? I use Apple Music” which added another strike on the app’s usability. Instead he got me thinking about using ChatGPT for the song recommendations instead.
And so began the second iteration: MoodTrackGPT, with planned features including: