Kangdi Wang
Portrait of Kangdi Wang

Kangdi Wang (Lertoon)

Audio AI Researcher & Engineer

Perceptual Audio Modeling · Music Generation · Audio Systems

I build audio models and tools with a focus on perceptual quality: how music preserves detail, phase, and stereo space. My work spans first-author research on εar-VAE and ear-VAE2, music generation and agentic audio understanding at Qwen Omni, and REAPER tooling through my open-source project Rea-Cli. Three years of mixing and live production inform both the problems I study and the way I evaluate results.

People always talk the talk, but not walk the walk.

Experience & CV Contact GitHub · DDDPG εar-LAB Google Scholar Research Intern · Qwen Omni
Contributor Audio agents

Qwen3.8-Omni: Towards Native Omni-Modal Agents

Qwen3.8-Omni-Flash

Contributed to Qwen3.8-Omni-Flash training, focusing on agentic audio understanding and REAPER-based tasks for automatic arrangement, mixing, and podcast editing.

arXiv
First Author Complex STFT

Fourier is Frontier: Frequency-Aware Autoencoding for High-Fidelity Music Reconstruction

ear-VAE2

A complex-spectral autoencoder that directly models frequency, phase, and stereo structure to address lost high frequencies and collapsed stereo images.

arXiv Code Project
First Author K-weighting

Back to Ear: Perceptually Driven High Fidelity Music Reconstruction

εar-VAE · INTERSPEECH 2026

Reconstructing music with perceptual weighting, phase supervision, and stereo-aware losses to preserve high-frequency detail and spatial cues.

arXiv Code Project
Core Contributor Foundation model

Qwen-Music Technical Report

Qwen-Music

Music generation from text and lyrics, plus cover-song generation. My work focuses on the audio decoder, listening evaluation, and quality-oriented data curation.

arXiv
Co-First Author Video dubbing

CineDub: Scaling End-to-End Video Dubbing to Multi-Speaker Dialogues with Coherent Sound Effects

CineDub · ACM MM 2026

An end-to-end video-dubbing model that generates multi-speaker dialogue and coherent sound effects from uncropped video.

arXiv Project
Co-author Semantic tokens

DuoTok: Source-Aware Dual-Track Music Tokenization for Vocal-Accompaniment Generation

DuoTok · ACM MM 2026

Separate vocal and accompaniment token streams preserve source structure for controllable music generation.

arXiv

Full publication list →

Rea-Cli

Open Source

REAPER Automation for Audio Agents

A Python library, CLI, and skill for audio agents. Edit RPP projects, run Lua/ReaScript in REAPER, and verify saved projects and rendered audio.

PythonCLILua / ReaScriptAudio Agents
GitHub

EAR-Audio-Preview

Open Source

Audio Analysis in VS Code

Inspect waveforms, spectrograms, phase, and loudness alongside model code. Built for listening to and diagnosing audio outputs without leaving the development workspace.

VS Code ExtensionTypeScriptWeb Audio APIOpen Source
GitHub

ReaperDoc

Live Demo

Interactive REAPER .RPP Specification & Parser Reference

An interactive reference for REAPER project-file structures and parameter meanings. Helps developers inspect tracks, items, and effect state when building DAW tooling.

DAW EngineeringReact / TypeScriptREAPER APIOnline Demo
GitHubLive Demo

ear-eq

Open Source

Reusable Interfaces for Audio Tools

Reusable knobs, EQ curves, and interaction components for audio software. Brings mixing-oriented controls into React and TypeScript interfaces.

ReactTypeScriptTailwindDSP UIDesign System
GitHubLive Demo

All tools & projects →