Logo
← Back

October 8, 2026

Introducing Drizel

A granular synthesizer project shaped by DSP, JUCE, and a long-term curiosity for electronic sound design.

The idea for a granular synthesizer wasn't new to me. I had been dreaming of building a studio with electronic gear mostly filled with synthesizers. My thought process is why not build something valuable and complex for a portfolio project through which I could learn a lot regarding music technology, especially digital signal processing and audio programming, with the JUCE framework?

Drizel logo

I had heard many songs earlier along the years where a granular synthesizer was used. A minimal definition of a granular synthesizer would be

A granular synthesizer plays a sound as many tiny overlapping fragments, called grains, typically 1-100 ms long. Each grain is a short slice of a sample shaped by a smooth window (fade in and out) so it doesn't click. Grains are triggered rapidly, and each can have its own start position, pitch, and length. Summed together, they form a continuous texture whose character comes from controlling those parameters: where in the sample grains are taken from, how dense they are, and how much they vary.

I'd been reading the lecture notes from George Mason University called 'Computational Music Synthesis' and C++ too, which would lay the basic foundation that would be required to build more blocks of this objective.

I aim to build a custom granular synthesizer where I'll be able to add features and technologies as it evolves through the development. I named it 'Drizel,' as the sound synthesized by a granular synthesizer sounds like a drizzle of rain. Very Subjective xD. A logo of rambutan is a bit open, as I wanted a red fruit on a light green background. Again, creativity.

On presenting this idea to a connection on LinkedIn whose name is Tero, he came up with an idea of LLM modulation, which I'm picking up in an experimental angle, sounds fun, and is definitely futuristic.

The code is licensed and open-sourced on GitHub. https://github.com/neuralmanacle/drizel

In the Read Me section you could find the architecture and game plan for the development and testing.

The major milestones are

  1. Set up the project: a clean JUCE and Projucer build with a first automated test.
  2. Load a sound: read a WAV file into memory and write it back out.
  3. Build the small pieces: window shapes and interpolation, and how each changes the sound.
  4. Play one grain: a short, faded slice at a chosen pitch.
  5. Play many grains: a scheduler that triggers overlapping grains with random variation.
  6. Run it live: move the engine into a real-time audio callback.
  7. Add modulation: LFOs and random walks so the sound evolves.
  8. Make it playable: MIDI input, an envelope, and presets.
  9. Build the interface: waveform display, controls, and a plugin build.
  10. Add AI control: a language model steering position and density slowly in the background.

Command Palette

Search for a command to run...