I Built an Open Source AI Music Plugin. Here's Why I Also Charge for It.

I Built an Open Source AI Music Plugin. Here's Why I Also Charge for It.

posted 5 min read

And why "open source = free labor" is killing independent developers


TL;DR for the impatient

  • I spent 500+ hours building OBSIDIAN Neural, a real-time AI music generation VST plugin
  • 100% of the code is open source (AGPL-3.0) on GitHub
  • I tried building community infrastructure first. Nobody helped.
  • I launched a cloud subscription service. People called me greedy.
  • This is my response.

The Project

OBSIDIAN Neural is the first VST3 plugin for real-time AI music generation designed for live performance. It's not a song generator like Suno or Udio. It's an instrument.

What it does:

  • 8-track MIDI-triggered sampler
  • Generates 30-second audio samples in ~10 seconds
  • LLM brain for contextual prompts
  • Draw-to-audio experimental mode
  • BPM sync, multi-output, step sequencer

I presented it as a Late Breaking Demo at AES AIMLA 2025 at Queen Mary University London.

The code? Fully open source.

You can clone it, fork it, and modify it. AGPL-3.0 license.


The Problem

The plugin uses Stable Audio Open for generation, which requires:

  • An NVIDIA GPU (CUDA-capable)
  • Python environment setup
  • Local server management
  • Technical know-how

Translation: Most bedroom producers can't use it.


The Community Experiment

Before launching any paid service, I did what every "good open source maintainer" is supposed to do:

I asked for help.

I opened Github discussion proposing:

Option 1: Community Cloud Partners

  • Seek sponsors willing to provide infrastructure
  • Freemium model with generous free tiers
  • Ethical pricing

Option 2: Peer-to-Peer Network

Inspired by LocalAI and Exo Labs:

  • Users with GPUs share compute resources
  • Community members contribute to a shared pool
  • Fair credit system for contributors

Option 3: Hybrid Approach

  • Local inference (if GPU available)
  • Community P2P network (shared resources)
  • Cloud fallback (minimal cost)

The Response

Crickets.

Zero cloud sponsors.
Zero P2P contributors.
Zero network engineers.
Zero DevOps volunteers.

Not a single person offered to help build the community infrastructure I proposed.


The Decision

I had three options:

  1. Abandon the project (let it die)
  2. Keep self-hosting only (lock out 90% of potential users)
  3. Build a paid cloud service (make it accessible)

I chose option 3.


The Cloud Service

I launched a subscription-based cloud API:

Pricing:

  • Starter: €14.99/month (500 credits)
  • Pro: €29.99/month (1500 credits)
  • Studio: €59.99/month (4000 credits)

What it covers:

  • Stable Audio API costs (charged per generation)
  • Gemini LLM API costs (prompt optimization)
  • Server infrastructure (hosting, bandwidth, SSL)
  • My time as a professional software developer

The Backlash

The moment I announced pricing, the comments rolled in:

"Greedy developer! Just another AI tax!"
"Subscriptions are evil! Make it free!"
"AI kills creativity! You're part of the problem!"
"Open source means FREE. Stop monetizing!"


Let's Address This

1. "Open Source = Free Labor" is a Lie

Open source means the code is accessible.
It does NOT mean I work for free forever.

Examples of successful open source + paid services:

  • Red Hat Linux → Open source OS, sold as enterprise service
  • GitLab → Open source code, paid cloud hosting
  • WordPress → Open source CMS, WordPress.com charges
  • Nextcloud → Open source software, paid hosting plans

This is the standard model. Nobody calls Red Hat "greedy."


2. "Just Make It Free!" — It IS Free

Want OBSIDIAN Neural for free?

Here's how:

  1. Clone the GitHub repo
  2. Install on your NVIDIA GPU
  3. Use it forever, no subscription

The code is yours.
I don't profit from your self-hosted setup.

The subscription is for people who:

  • Don't have GPUs
  • Don't want to manage servers
  • Want convenience

3. "You Should Build Community Infrastructure!" — I Tried

I spent months proposing community solutions.

Nobody contributed.

So I built something sustainable that lets non-GPU users participate.


4. "Software Development is My Job"

I'm a professional software developer.

I spent 500+ hours building OBSIDIAN Neural:

  • Initial development
  • Bug fixes
  • Cross-platform compatibility (Windows, macOS, Linux)
  • VST3 + AU formats
  • Documentation
  • User support

Being compensated for professional work isn't greed. It's called having a job.

Would you work 500 hours for free?


The "AI Kills Creativity" Argument

Let's be clear:

OBSIDIAN Neural does NOT replace musicians.

It generates raw material.
YOU arrange it.
YOU mix it.
YOU perform it.
YOU shape it into music.

It's like saying:

  • "Synthesizers kill creativity because they generate waveforms"
  • "Samplers kill creativity because they play pre-recorded sounds"
  • "DAWs kill creativity because they automate recording"

Every tool automates something.
Creativity is HOW you use it.

If you think AI-generated samples are "lazy," fine-tune your prompts or your models, experiment with draw-to-audio, explore edge cases.

Use your brain.
That's where creativity lives.


The Numbers

Since launching:

  • 150+ GitHub stars
  • Press coverage: Synthtopia, KVR Audio, Bedroom Producers Blog, AudioFanzine, MIDIFAN (China), S1 Forum (Korea), FutureMusic España, Rekkerd, DTM Plugin Sale (Japan)
  • Featured at AES AIMLA 2025 (Audio Engineering Society conference, London)

The cloud service is working.
People who want convenience are paying.
People who want free self-hosting are doing that.

Both are valid.


What I've Learned

1. "AI Haters" Are Loud, But Irrelevant

The people screaming "AI kills creativity" aren't your users.

They never were.
They never will be.

Your users are the ones quietly building with your tools.


2. You Can't Please Everyone

Steve Jobs:

"If you want to make everyone happy, don't be a leader. Sell ice cream."

Build for the people who get it.
Ignore the rest.


The Bottom Line

I built a tool I'm proud of.

I released the code for free.

I explored every community option before launching a paid service.

Nobody helped.

So I built something sustainable.

If you think that's "greedy," you're welcome to:

  • Self-host for free
  • Fork the code and build your own service
  • Contribute to the community infrastructure I proposed

But don't tell me I should work for free while you profit from my labor.


Final Thoughts

Open source is beautiful.

But open source ≠ free labor.

Developers deserve to be compensated for their work.

If you disagree, build it yourself.

The code is right there.


OBSIDIAN Neural:
obsidian-neural.com
GitHub
Download v216 (VST3 + AU)
AES AIMLA 2025 Paper

1 Comment

1 vote

More Posts

How I Built a React Portfolio in 7 Days That Landed ₹1.2L in Freelance Work

Dharanidharan - Feb 9

AI Reliability Gap: Why Large Language Models are not for Safety-Critical Systems

praneeth - Mar 31

I’m a Senior Dev and I’ve Forgotten How to Think Without a Prompt

Karol Modelskiverified - Mar 19

I built a cooperative GPU network for real-time AI music generation — looking for providers

Innermost47 - Mar 30

Most Startups Add AI Too Early — Here’s How I Decide When It’s Worth It

kajolshah - Jan 8
chevron_left

Related Jobs

View all jobs →

Commenters (This Week)

2 comments
1 comment

Contribute meaningful comments to climb the leaderboard and earn badges!