Skip to main content
2026-04-14
Feature

Spring 2026 Update

It’s been a busy start to the year. Since January, I’ve been heads-down building out features that I’ve wanted for a long time - and a few that surprised me along the way. Here’s what’s new.Cover cards in user collection

Your Collection, Your Way

The biggest addition this spring is the new Collection & Wishlist system. This has been the most requested feature since FMDB launched, and I’m happy to say it’s finally here.You can now track which soundtracks you own, which ones you’re eyeing, and browse your collection the way you’d browse a record shelf. Every release page has 2 buttons that lets you add it to your collection or wishlist with a single click - no more navigating away to a separate page.Here’s what’s included:A dedicated Collection page where you can browse, filter, and search everything you’ve added. Grid view with covers, faceted search by format, label, year - it works the way you’d expect.Privacy controls. A three-tier visibility system (private, friends, public) lets you decide who sees your collection. This is a global setting - simple and predictable.Collection statistics that give you a quick overview: how many releases, which formats, which labels appear most. Nothing fancy, but it’s satisfying to see the numbers.Quick-add buttons across the site. When you’re browsing search results or an artist page, you can add releases to your collection without ever opening the detail page.The collection system is available to all registered users.See all features: https://www.fmdb.net/features/collection-management

An API Is Taking Shape

Behind the scenes, I’ve been building out a proper REST API. It’s not publicly available yet - right now it powers the upcoming iOS app and some internal tooling - but it’s growing fast.The API covers releases, artists, labels, productions, recordings, collections, wishlists, and more. It uses JWT authentication, follows OpenAPI conventions, and returns properly structured data including cover images in multiple sizes via imgproxy.Some of the more specialized endpoints include audio feature submission (timbre, harmony, texture vectors per track), audio fingerprints, IFPI label codes, and disc scan data. These are building blocks for features that will become visible to users over time.When the API reaches a stable state, I plan to open it up - first as a beta for interested developers, eventually as a core part of the platform.

Finding Things Faster

Two additions that have changed how I use FMDB myself:Quick Search (⌘K). Press Cmd+K (or Ctrl+K) anywhere on the site and a command palette opens. It searches across releases, artists, labels, and productions simultaneously via Meilisearch’s multi-index search. Results appear as you type. I use this dozens of times a day now - it’s the fastest way to navigate the database.Quick SearchBarcode Scanner. On mobile, the search page now has a camera icon that opens a barcode scanner. Point it at a CD’s EAN/UPC barcode and FMDB will find matching releases instantly. This started as a “wouldn’t it be cool if…” idea and turned out to be genuinely useful when cataloging physical media.On top of that, search results now show human-readable facet labels instead of internal codes. Country names instead of ISO codes, proper type labels instead of database enums. Packaging types (jewel case, digipak, gatefold…) are now a searchable facet too, along with a physical/digital medium filter. Small touches, but they make browsing feel a lot more natural.

Source Cues on Tracks

A small but meaningful addition for film music nerds: tracks can now be marked as source cues. These appear as amber “Source” badges in the tracklist view.If you’re not familiar with the term - a source cue is music that exists within the world of the film. A song playing on a radio, a band performing at a party. It’s distinct from the underscore, which is the “invisible” music the audience hears but the characters don’t. Being able to distinguish the two in a tracklist gives you a much better picture of how a score is structured.
Releases, artists and labels can now be linked to their MusicBrainz and VGMdb pages. These links appear on detail pages and help connect FMDB’s data with two of the most important external databases in the music and soundtrack space.MusicBrainz is the open music encyclopedia - it’s especially strong on metadata standards and identifiers. VGMdb is the reference for video game and anime music, an area where FMDB is growing but VGMdb has decades of depth.Apple Music badges appear on production pages where applicable.

Schema.org: Better Visibility in Search Engines

This one isn’t visible in the UI, but it matters for discoverability. Release pages now include Schema.org structured data in JSON-LD format.Every release page outputs a MusicAlbum schema with track listings, artist credits, label info, and ISRC codes where available. Artist pages output MusicGroup schemas. The site also supports BreadcrumbList navigation and a SearchAction so Google can offer direct searches into FMDB.The result: FMDB pages are now eligible for rich results in Google Search - album cards with track listings, artist knowledge panels, and direct search integration.

Platform Statistics

There’s a new Statistics page that shows a high-level overview of the database: how many releases, artists, labels, and productions are catalogued. It’s a simple snapshot, but it’s the first time this data has been surfaced publicly.

Behind the Scenes: The Release Backlog

This section is a bit different - it’s about a system that most users won’t interact with directly, but that shapes everything they see in the database.The Release Backlog is where new releases enter FMDB before they become full database entries. When I discover a new soundtrack release - from a press announcement, a label catalog, a forum post - it goes into the backlog as a raw text line. Something like:
Jerry Goldsmith – Logan's Run [FSMCD Vol.5 No.2 | 2002-02 | 1 disc | Silver Age Classics | Label: Film Score Monthly]
The old system tried to use AI to automatically parse and match these entries against the database. It didn’t work well. Matching was unreliable, false positives created duplicate entries, and I spent more time fixing AI mistakes than I saved.So I rebuilt the entire pipeline from scratch.A deterministic parser that understands the backlog format - artist names, titles, production links in parentheses, years, media formats, catalog numbers in brackets. No AI, no guessing. It either parses correctly or flags what it can’t handle.A matching engine that checks parsed results against existing releases, artists, and labels in the database using the search. It surfaces potential matches with confidence scores so I can review them quickly.A review UI with a slide-over panel, smart queue tabs (unmatched, matched, review needed), batch actions, and real-time updates. The workflow went from “open entry, click through five pages, come back” to “scan, confirm, next.”User proposals. Registered users can now suggest releases that should be added to the database. These feed into the same backlog pipeline, so nothing gets lost and everything follows the same review process.The old AI pipeline has been completely removed. The new system is faster, more predictable, and actually trustworthy. It’s a good reminder that the right boring solution usually beats the exciting fragile one.Releases Backlog

What’s Next

There’s plenty more in the pipeline. The main focus is on adding releases to the database. Thanks for reading, and as always - if you have feedback, hit the button in the bottom right corner. I read everything.— Marcus