Show HN: Unheard.fm – No-repeat, user-controlled music discovery for Spotify
My name is Max. After a sabbatical from the corporate world, I turned a passion project into a full web app: Unheard.FM.
What is it? Unheard.FM is a music discovery web app for Spotify that selects tracks based on your direct preferences instead of listening habits, and provides fresh results by excluding your "known" tracks. (No paid sponsors here.)
Why did I build this? A few reasons... 1. I was getting tired of getting too similar or repeat recommendations from Spotify and thus wasn't getting out of my music "comfort zone" enough. 2. I was looking for ways to source more tracks of specific types of music and didn't just want "popular" music. 3. I listen to a lot of music and didn't want to wait for new curated playlists to drop.
How is it different from other music discovery apps? * Your discovery model is completely transparent and controllable by you * Not based on Spotify's recommendation API or sponsorships. The engine simply: - Selects random sets of tracks by provided seeds (years, genres, desired popularity distribution) - Filters by provided excludes (artists, genres, track attributes, etc) * Results are always fresh since your liked, playlist, top, and recently played tracks are filtered automatically * Local-first: Known tracks are cached on your device; no listening history is stored server-side.
Backstory This started as a Python CLI. I was in a competition with my (now) wife to create an ultimate playlist for a random year (I had 1962). I had issues using the Spotify client to get enough year-based content, so I wrote a CLI to fetch a random set of tracks for a year (Spotify API). Random year-based discovery turned out to be surprisingly fun (and often hilarious), so I kept adding features...Then I wanted to use it from anywhere, so I built a quick static react web app (literally a single page form) and hosted it via a S3 bucket... many years later it grew into Unheard.FM.
The Stack React frontend + serverless backend (AWS + FastAPI). Most discovery logic runs client-side; the backend mainly mediates Spotify auth and serves lightweight discovery APIs.
What's Next? This project still has a long way to go!... but a few things on my roadmap include:
* Regional Filtering (by far the biggest user ask thus far, but also the most difficult to execute on) * More UX improvements (additional filters, better nav, optimized desktop view) * Multiple "discovery models" / sharing * Support of other music platforms (Apple Music, YouTube) * Exploring a potential paid tier to help fund development and cover costs, which might include: Multi config, advanced filters, bigger playlists / feeds.
Hope y'all have some fun with it. Any and all feedback is much appreciated!
No comments yet