Music streaming platforms all claim to use both artificial intelligence and manual curation to find new songs from emerging artists. But users often have to listen to many songs to find some likeable songs – that’s because they have no control over the recommendation algorithm. Romanian developer duo Alex Ruber and Andrei Patru have developed an app called Smores that improves this process and helps you easily add new music to your library.
Smores is a free iOS app which allows users to listen to a short clip of a song based on their listening history. Just like TikTok, users can browse the songs using a vertical feed.
The app connects to your Spotify account and uses the Spotify API to find new songs for you. If you enjoy a song clip, you can tap the like button and it will be added to a playlist called “Smores discovery” in your Spotify account. Alternatively, you can also add the song to one or more of your pre-existing playlists.
The developers told TechCrunch in an email that they were going to build the app to discover new music on their own. So they launched the first version of Smores last September.
“We love discovering new music, but we were trapped in our recommendation bubbles and it took us too long to sift through the huge amount of new music coming out. At the same time, we had a hunch that you only need to listen to the ‘right’ snippet of a song to know whether you like it or not: Shazam’s popularity indicates that this is the case,” they said.
The duo said they wanted more control over the discovery algorithm and built transparency into the app. To that end, Smores has a ton of built-in controls to change users’ recommendation feed. Users can filter suggestions based on their top six microgenres of the month. These change as they listen to more music and like more songs in the app.
The app’s advanced settings allow you to define the clip length (from 5 seconds to 60 seconds); limit discovery based on how many followers an artist has on Spotify; and filter songs by BPM (beats per minute), song key, and release date.
One of the great things about the app is that it ensures that you will never listen to the same song again. Additionally, the developers said they’ve tweaked the algorithm so that it determines the “best” part of the song to play in a clip. They said many users like the song after just five seconds if the app plays the right preview.
Retention of users and future plans
Music discovery apps are fun to use, but it’s hard to build an audience that regularly uses an app. Despite this challenge, the developers said they managed to maintain a high number of users (7% for week eight) and have heard positive things from regular users.
“It is true that the cadence is low for music discovery apps in general. The casual listener will actively discover new music, perhaps once every three months. Casual listeners, DJs and playlist managers rave about how much they love the ease of use, speed, convenience and quality of our recommendations,” they said.
Currently, the team is focused on building features like Smores radio and integrating Apple Music or other streaming platforms. Eventually, they want to introduce an Android version and possibly a premium tier, though they haven’t committed to the paid features yet.
More AI in music
Music fans have often complained about the growing role of AI in music discovery and distribution. And yet, companies and app developers rely more on AI, but they use it to give more control over algorithms with buttons and filters.
Bytedance’s music app Resso — currently only available in India, Brazil, and Indonesia — banks on a vertical feed and the company’s proven AI prowess at helping casual listeners find new artists. The Chinese tech giant is also aiming to launch TikTok Music globally — AI-powered music suggestions are likely to play a vital role in the service.
App developers are also taking the help of AI to introduce features in music apps. LineupSupply, an app that turns festival posters into playlists, changed its name to Playlist AI. The app has also been introduced a new function which lets you write a prompt like “Dance artists who were popular in the 90’s” to generate a playlist.