Building Psychill Space: A data-driven archive, real-time genre mapping, and community platform for psychedelic, electronic and experimental music

[varkus]

Hey everyone,

I’m a music producer and software developer, and for the past year, I’ve been building Psychill Space - a web platform designed for discovering and sharing music, and connecting with like-minded people.

Here is a breakdown of the core features and the logic behind them:

Relational Music Archive & Crowdsourced Metadata

At the core of the platform is a relational database containing hundreds of albums across over 100 micro-genres.

• Dynamic Track Profiling: Users can vote on specific track genres, moods, and tags. This crowdsourced metadata continuously refines the data model.
• Real-Time Data Visualization (Genre Map): As users profile tracks, a dynamic 2D Genre Map updates in real time. It uses spatial clustering to display similar and coexisting genres next to each other, allowing users to visually track which albums gravitate toward specific sonic nodes.
• Advanced Querying & Discovery: Built-in advanced filtering and a personalized recommendation engine (Discovery Mode) that utilizes the crowdsourced tag matrices.

Platform-Agnostic Media Aggregation

One of the main UX challenges I wanted to solve was fragmented music streaming links.

• Unified Track Architecture: Every track in the database acts as a central hub containing links to Bandcamp, Spotify, Apple Music, Tidal, and more.
• Seamless Cross-Platform Sharing: Users can share music within the community feed, and recipients can stream it on their preferred service with a single click - no manual searching required.

Community & Events

• The Feed/Forum Hybrid: The social layer is structured as a hybrid between a real-time activity feed (similar to Twitter/X) and a nested relational discussion board (similar to Reddit/forums), built completely ad-free.
• Event & Lineup Mapping: The events section acts as a directory for festivals and gatherings. It maps lineups directly to the artists' discographies within our database, allowing users to query local event nodes based on specific genre filters. (The database is still actively being populated).

To build this, I'm using supabase, nuxt, and various open-source projects/libraries. I've used LLMs to help me with the project, but the entirety of the content is human-made.

I'd love to hear your thoughts on the platform's UI/UX, the genre-mapping visualization, or any suggestions you have for handling crowdsourced music metadata!

 

[Nydex]

Hello Varkus,

The idea for your platform is technically not bad, but I want to poke it a bit and see how it actually works under the hood before we can consider allowing you to advertise it here.

My main concern with what you've created is the ethics. As you know in this day and age of actual artists being basically robbed by AI companies, to even begin contemplating the promotion of a platform like this, we need to establish where your ethical foundation lays and how it looks like. A few questions related to that:

1. You claim the platform solves the "fragmented music streaming links" problem, but the exact mechanics of playback dictate whether artists are paid fairly. Does your platform host or cache any audio files natively, or does it strictly function as a directory of external links? If it uses embedded web players (like the Spotify or Bandcamp iframe), do those plays reliably count toward the artist's monetized stream count on the native platforms?
2. While the platform is currently ad-free and doesn't require a subscription (at least from what I've seen so far, which admittedly isn't much), what is the long-term monetization strategy? If the platform introduces premium features, affiliate links, or data-selling in the future, will the artists whose work generates the platform's traffic receive any share of that revenue?
3. If your platform servers strictly as a routing agent, when a user clicks to play a track, does the platform prioritize routing users to high-payout platforms (like Bandcamp) over low-payout platforms (like Spotify)?

Next, a few questions on data sourcing and your recommendation engine. The platform utilizes a "relational database" and a "personalized recommendation engine." It is important to establish how the initial dataset was built and what the engine actually analyzes.

1. You mention hundreds of albums are already in the database. Were these scraped from existing platforms like Discogs, Last.fm, or the Spotify API? If so, does the platform comply with their respective API terms of service regarding commercial use, rate limits, and data storage?
2. Does your spatial clustering algorithm only process user-submitted text tags, or does it ingest and analyze the actual copyrighted audio files to determine "sonic nodes"? Ingesting and analyzing copyrighted audio without a license is a pretty serious legal issue, for obvious reasons.
3. You mentioned using LLMs to help with the project. Were LLMs used to scrape, generate, or summarize artist biographies and album descriptions? If so, how is the platform ensuring accuracy and preventing the generation of misleading information about independent artists? At the very bottom of your website (which is incredibly laggy by the way) it says this:
   
   But have you actually contacted any of the artists whose work is on your platform and gotten their legal agreement for their music to be put on there, and basically bound them to be the ones responsible for maintaining the validity of the information in their profiles?

And finally a couple of questions about the data ownership and opt-out mechanisms. Your platform seems to be designed in such a way that it relies heavily on users voting on genres and moods to refine the data model. So:

1. Since the community is providing the labor to build and refine the "crowdsourced tag matrices," who owns this resulting dataset? Is the database open-source and exportable for the community, or are you locking community-generated value into a proprietary, closed system?
2. Do artists have a clear, accessible way to opt out and have their discographies, metadata, and associated event listings completely removed from the platform if they do not wish to be included in this specific ecosystem?

Please understand that we are a platform that holds integrity very high and we will not allow something that is inherently immoral, especially towards artists, to be promoted hereon. Furthermore, you can take these questions as something that can help you, because if you haven't properly thought out some of these points, you can get into some considerable legal trouble. It's happened before, and for a very good reason.

Looking forward to your answers.

 

[varkus]

You claim the platform solves the "fragmented music streaming links" problem, but the exact mechanics of playback dictate whether artists are paid fairly. Does your platform host or cache any audio files natively, or does it strictly function as a directory of external links? If it uses embedded web players (like the Spotify or Bandcamp iframe), do those plays reliably count toward the artist's monetized stream count on the native platforms?

If your platform servers strictly as a routing agent, when a user clicks to play a track, does the platform prioritize routing users to high-payout platforms (like Bandcamp) over low-payout platforms (like Spotify)?

While I allow adding Spotify links, I actually openly criticize the platform and encourage to purchase directly from Bandcamp. As of now, the platform does not host or cache any audio files. I'm currently working on a way for artists to *voluntarily* upload either short previews or full tracks (after they prove their ownership) - could be working in next days. Where possible, I use Bandcamp embeds - the streaming quality is capped at mp3 128kbps, and artists actually can see in their dashboard how many plays are they getting, where from, and if they don't like it, they can disable it with one click. After long considerations, I think this option is ideal. I avoid embeds that use cookies.

While the platform is currently ad-free and doesn't require a subscription (at least from what I've seen so far, which admittedly isn't much), what is the long-term monetization strategy? If the platform introduces premium features, affiliate links, or data-selling in the future,

The platform is and will stay ad-free and won't require a subscription. I practice data minimization, so there won't ever be any behavioral data to sell. Affiliate links aren't off the table, but I would only use links that genuinely match the site's topics and they would be clearly marked. I plan to add a low-cost subscription which would include cosmetic features, and maybe stickers if I get enough traffic and find artists to collaborate with.

will the artists whose work generates the platform's traffic receive any share of that revenue?

That's not right. It's my blog posts and my work that drive the traffic. And I put the spotlight on the producers, and redirect interested users to their Bandcamp pages.

You mention hundreds of albums are already in the database. Were these scraped from existing platforms like Discogs, Last.fm, or the Spotify API? If so, does the platform comply with their respective API terms of service regarding commercial use, rate limits, and data storage?

The data was imported manually and with help of Musicbrainz API/Discogs API, which were made specifically for such cases. The data is CC0, so it is free to use for commercial projects. If I weren't abiding by rate limits, I would be getting 429.

Does your spatial clustering algorithm only process user-submitted text tags, or does it ingest and analyze the actual copyrighted audio files to determine "sonic nodes"? Ingesting and analyzing copyrighted audio without a license is a pretty serious legal issue, for obvious reasons.

Analyzing copyrighted audio could be considered fair use actually, though for a potentially commercial project it would be disputable. And no, I don't do it, the algorithm depends entirely on user-submitted genre and mood sliders, and text-based tags.

You mentioned using LLMs to help with the project. Were LLMs used to scrape, generate, or summarize artist biographies and album descriptions? If so, how is the platform ensuring accuracy and preventing the generation of misleading information about independent artists? At the very bottom of your website (which is incredibly laggy by the way) it says this:

No, LLMs were mainly used for brainstorming, architecture and boilerplate. I didn't generate any discographies or other content.

which is incredibly laggy by the way

Do you mean it loads slow, or you experience graphical bugs, or do you see skipping frames or something like that? If it's the latter, would you mind sharing your browser and OS?

But have you actually contacted any of the artists whose work is on your platform and gotten their legal agreement for their music to be put on there, and basically bound them to be the ones responsible for maintaining the validity of the information in their profiles?

Do artists have a clear, accessible way to opt out and have their discographies, metadata, and associated event listings completely removed from the platform if they do not wish to be included in this specific ecosystem?

I don't need legal agreement to put an embed or streaming links on a website... But several artists have claimed their profiles and submitted their music to the platform. If anyone doesn't like their music being there, I'm letting them make takedown/opt-out requests, even though I don't really see a reason why anyone would like to do that.

Since the community is providing the labor to build and refine the "crowdsourced tag matrices," who owns this resulting dataset? Is the database open-source and exportable for the community, or are you locking community-generated value into a proprietary, closed system?

No, currently users can only export their own data. Open-sourcing the dataset is definitely a possibility I could consider, though I'm afraid that if I did, it would only be used for bad purposes and none for good.