- Spotify's iOS app automatically generates a background glow around the album art of the currently playing song. I first became interested in Spotify's implementation of this feature when I noticed that the glow color of Brockhampton's "Saturation III" was different for different songs despite them sharing the same album art. (as of 09/07/2019 Spotify has fixed this bug)
- This is somehow related to the colors present in the artwork, but not necessarily an exact replication of the art (like android's scale and blurred art when playing)
- Sometimes the glow is highlight/accent colors of the artwork, but other times it is just the most prevalent color.
- This project is my attempt to write code that generates a glow color value from colors in an album cover.
- Images:
- 9 Screenshots of random album art from my phone, including 2 songs on the same album, which share the same art, but have different glow colors.
- 8 Images of album art from the screenshots.
- Code:
- spotify_album_blur.md the README file for this project.
- requirements.txt environment packages used in development.
- album_blur.ipynb jupyter notebook with the actual code.
Last update: September 9, 2019
- I am happy with the functionality of the code as it stands here. It was a good hand-on example of image-processing techniques and thinking of how a large company would implement such an easy to overlook feature.
- Create mock-up UI from the extracted highlight colors. (Seems relatively straightforward and not really in my area of interest)
- Fine-tune image scaling and k-means clusters parameters.
- These have a large impact on which colors are produced. I noticed while testing that colors would be removed during scaling and clusters would move to non-existent colors while calculating means.
- Look into other ways of determining the most prominent color.