All Podcast

Podcast

How to Build Audio Plugins With a Full-Time Job w/ Simon Weis (Dark Palace Studios) | Ep 41

Simon Weis of Dark Palace Studios on building audio plugins as a side business: pricing, no DRM, custom OpenGL UIs, JUCE, and where AI really helps.

Three Things You Can Take Away From This Episode

  • Treating a plugin company as a side business removes the financial pressure that pushes people to ship before the product actually feels right.
  • For audio DSP in C++ and JUCE, today's LLMs are roughly 80% miss and 20% hit: great for the boring parts, unreliable for the hard ones.
  • In a saturated plugin market, a standout UI, genuinely unique effects, and an email list you actually own are what cut through the noise.

The Audio Programmer Podcast welcomed Simon Weis, the developer behind Dark Palace Studios, joining from near Nuremberg, Germany. Simon has a PhD in physics and works on automotive and CAD software by day, then builds audio plugins at night, a routine he calls moonlighting.

Dark Palace Studios is a small, deliberately part-time operation. Simon handles the DSP and code, Kenneth designs the UI (he previously worked on Vital's interface), and Maxime runs marketing and social. Their plugins are named after animals, priced low on principle, shipped without DRM, and built with a custom OpenGL renderer instead of stock JUCE graphics. The conversation moved from how the label started to pricing philosophy, the realities of using AI for audio code, and why neither Simon nor Josh believes "software is over."

1. A Side Business Can Afford to Put Quality First

None of the Dark Palace team works on plugins full time, and Simon sees that as an advantage. Because nobody needs the revenue to pay their bills, there is no pressure to rush a release to hit a number. That freedom changes the product itself.

When something is not ready, they let it cook. Building the DSP is only the first few weeks of work. Getting the UI, the presets, and the overall feel right takes months, and that is the part that decides whether people love the plugin or forget it. A team under financial pressure rarely gets to protect that time.

2. Faster Tools Still Need Someone Who Understands the Code

Simon credited the recent wave of LLM coding tools with lowering the barrier to building plugins, and he is glad more people can get started. He was also honest about the downside. If you do not understand or even read the code an AI produces, you cannot judge it, and the symptoms show up fast: plugins that eat CPU because nothing is optimized, obvious graphics glitches, and behavior the builder cannot explain.

His point was simple. Going faster only helps if you understand what the code is doing. Otherwise you reach the wrong place sooner.

3. For Audio DSP, AI Is Still 80% Miss and 20% Hit

Simon experiments with AI tools regularly, and his hit rate for audio work is specific: about 20%. For repetitive chores (renaming things, converting a JSON file from camelCase to lowercase, porting a change from one plugin to another) it is genuinely useful. For the hard problems, he says it produces garbage: wrong upsampling and downsampling, broken parameter smoothing, and allocations on the audio thread, which is exactly what you must never do.

His theory is that the specific combination of C++, JUCE, and audio DSP is thinly represented in public repositories, so the models never had enough training data. Ask the same tools to transform some Python and the output is far stronger. The lesson for audio developers is to calibrate your expectations by domain rather than assuming the tool is equally capable everywhere.

4. Write by Hand What You Care About, Delegate What You Do Not

Both Simon and Josh landed on the same working rule, one Josh credited to his friend Sudara: write by hand the things you genuinely care about, and let AI handle the things you do not. For Simon that means writing the DSP himself, because that is the craft he enjoys, while happily handing off CI scripts and installers.

Part of why that split works is structural. Installers and CI scripts tend to be binary. They either run or they do not, and they have few cross-dependencies. UI and DSP code are full of soft requirements and dependency chains: a bypass that has to gray out the right controls, code that must not allocate, run within a time budget, and keep latency low. Those are the places where a plausible-looking answer is most likely to be quietly wrong.

5. Low, Fixed Pricing on Principle

Dark Palace sells most plugins for around €20, at a single fixed price, and has not raised prices since 2023. There is no Black Friday 90% cut, by design. Simon's view is that constant deep discounting trains customers to wait and pressures them with artificial urgency, and he would rather let people make an informed decision in their own time.

He was candid that not needing the income makes this easier, and that bigger companies with real overhead cannot move as lean. The goal is sincere: giving people who are not as financially comfortable, including many outside wealthy Western markets, access to good tools without resorting to cracks.

6. No DRM, Because It Only Punishes the Honest User

Simon ships no DRM at all, for two reasons. He does not enjoy building it, and he does not think it works. Anyone determined to crack a plugin will. He has been on the receiving end himself, once tampering with his PC's MAC address to recover a license after a machine died, with clients waiting to record.

DRM is never fair to the honest user, he says, and the honest users are the ones who were going to pay anyway. Dark Palace also ships Linux builds for every plugin, and Simon has seen real growth and loyalty from the Linux community. He is a Linux user himself.

7. Cut Through the Noise With UI, Unique Effects, and a List You Own

Asked how a small label gets heard in a crowded market, Simon pointed to three things. First, a UI people want to touch and drag; Kenneth's design work does most of that. Second, effects you cannot get elsewhere, rather than another EQ competing with 500 existing ones. Third, and most important, your own newsletter.

He was emphatic about that last point. A forum, an Instagram following, or any other platform belongs to someone else who can cut you off at any time. An email list is yours. Free plugins offered in exchange for an email, combined with the occasional YouTube feature, have driven most of Dark Palace's growth.

8. Winning at KVR Was Good Engineering and Good Timing

Dark Palace took third place in the KVR Developer Challenge with Parrot, a multimodal delay pitched by Maxime. The signal splits into 3 modes, each either full-band or a narrow slice of the spectrum, with an independent delay per mode. Each mode gets its own character: tube distortion, an all-pass suite, or chorus, plus pitch and frequency shifting. It even key-tracks from a MIDI keyboard.

The visibility from the challenge mattered as much as the plugin. Simon saw a clear spike in traffic across the rating period and especially on results day, and a top-3 finish pulled new visitors toward the rest of the catalog and the full bundle. Credit where it is due: Parrot's pitch and frequency shifting come from Signalsmith's permissively licensed libraries, and Signalsmith took first place in the same challenge with Scintillate.

9. Custom OpenGL and a Clear-Eyed View of JUCE

Dark Palace renders its interfaces with custom OpenGL instead of stock JUCE graphics. The team runs high-refresh monitors and wanted their UIs to move at the screen's native rate, which JUCE's graphics could not deliver at the time. Simon rewrote the rendering layer, keeping JUCE's mouse and click handling but drawing everything himself. He considered Visage, but owning code he already understands beat adopting a new framework he would have to learn.

He has mixed feelings about JUCE overall, with one clear exception: shipping a single codebase to VST3, AU, CLAP, and LV2 across Windows, Mac, and Linux just works, and that saves him enormous time. CLAP support arriving in JUCE 9 is on his list to try.

10. Why Simon Is Optimistic About What Comes Next

Simon is optimistic about where the tools are heading. He compared this moment to music production in the 2000s, when cheap audio interfaces broke the old model of renting an expensive studio and made whole genres possible, synthwave among them. Josh offered a food analogy: grocery stores did not kill restaurants, because people still value being served and experiencing someone else's creativity.

Simon's final note was that when Napster and Spotify arrived, people declared the music industry dead. It is very much alive; it changed. His hope is that developers embrace that change rather than fight it. The joy of cracking a hard problem, or discovering a new effect from a stray sign error, is something he does not believe AI can reproduce.

Closing

Simon is not dismissing AI tools or pretending the market is easy. He shows what a sustainable, principled plugin business looks like: protect the time to make things feel right, use AI where it helps and write the rest by hand, price and license in a way you can defend, and grow an audience you own. For audio developers wondering whether there is still room to build, Dark Palace is a quiet argument that there is.

Business
JUCE
AI
Digital Signal Processing
C++
Music Tech
SHARE THIS

More Podcasts

View All
The Audio Programmer Podcast cover art
The Audio Programmer Podcast cover art
The Audio Programmer Podcast cover art
View All