A monophonic synthesizer with MIDI input, an interpolated delay effect, a lo-fi style bit crusher, and a low/high pass filter.
< Read More >
Implementation of a Schroeder Reverb (4 all-pass filters into 8 comb filters) in JUCE.
< Read More >
Compressure
A modulated delay chorus effect. < Read More >
A Lookahead compressor with a Soft/Hard knee toggle. This was my entry to the AES New York 2023 MATLAB Audio Plugin Hackathon. < Code >
A polyphonic synthesizer with reverb, delay, bit crusher, filters, and an envelope.
< Read More >
Developed a Wind synthesizer for the Unity game engine using the JUCE framework. < Code >
Audio Software/Plugins/Code
A pitch mod for the co-op game Lethal Company, made in C# using ILSpy, BepInEx pack, and the Harmony library to post-fix the game's underlying code.
Performed an acoustical analysis of a room using an exponential sine sweep and recording the room response using a measurement microphone, and proposed a potential room treatment based on the findings to reduce RT60 from 0.34 to 0.30, treat modal resonances, and reduce peaks at unwanted frequencies. Measurements were done with Room EQ Wizard software.
-----------------------------
-----------------------------
MUSHRA Test & Analysis
Performed a MUSHRA listening test for subjective assessments of perceived audio quality between audio files of varying bit depths, and used the ANOVA test to infer statistical differences between them.
Pure Data Projects
A procedural soundscape in PureData.
A procedural soundscape in PureData with extra steps [Using serial data communication to make Arduino talk to PureData for controlling parameters].
-----------------------------
Teaming up with students, made a PureData + Arduino arpeggiator that responds to shadows and light. We call it the LuminoSynth.
Machine Learning
A two-part project for MMI 611, Music & AI class.
Music Genre Classification (Part 1)
The task is to choose a machine learning task of choice, and then using the code from Sebastian Raschka's book on Machine Learning with PyTorch & Sci-kit learn, choose a neural network to complete the task.
This short report is my progress on part 1 of the project - choosing and training a smile classification convolutional neural network. [Read More]
Voice Activity Detection in Noise
The goal of this project was to understand how training a neural network works, and try to increase its accuracy through implementation of MFCCs or adding and removing layers such as LSTMs. [Read More]
OpenAI Assistant
Mixtape-Me is a function-calling OpenAI Assistant that uses the Brightdata web scraping browser (with puppeteer library) to return YouTube mixes with a specified keyword, such as a music genre.
< Github >