Hi, I am Wen-Yi Hsiao (蕭文逸).
I am a Research Scientist at ByteDance, Seed Team@San Jose, USA. 🇺🇸
I’m a researcher (1K+ citations) and programmer (1K+ GitHub stars), working to bridge the gap between theory and practice.
I work with teams to do research, create products, and organize exhibitions.
Meanwhile, I study and perform music on my own.
I love to work with people from diverse fields, where I believe true sparks happen.
I am also a piano lover 🐈.
2022 ICASSP: “Towards automatic transcription of polyphonic electric guitar music” Analysis - Arxiv | Dataset (2nd Author)
2022 ISMIR: “DDSP-based singing vocoders: A new subtractive-based synthesizer and a comprehensive evaluation.” Vocoder - Arxiv | Demo | GitHub (co-1st Author)
2021 EUSIPCO: “Source separation-based data augmentation for improved joint beat and downbeat tracking.” Analysis - Arxiv | GitHub (3rd Author)
2021 AAAI: “Compound word transformer: Learning to compose full-song music over dynamic directed hypergraphs.” AI Music - Arxiv | GitHub | Slides (1st Author)
2020 ISMIR: “Automatic composition of guitar tabs by transformers and groove modeling.” AI Music - Arxiv | Demo | Slides (3rd Author)
2020 MMSP: “Mixing-specific data augmentation techniques for improved blind violin/piano source separation.” Separation - Arxiv | GitHub (2nd Author)
2018 AAAI: “Musegan: Multi-track sequential generative adversarial networks for symbolic music generation and accompaniment.” AI Music - Arxiv | GitHub (co-1st Author)
Journals
2021 JNMR: “Automatic melody harmonization with triad chords: A comparative study.” AI Music - Arxiv | GitHub (2nd Author)
Workshops
2019 ISMIR Late-Breaking: “Jamming with yating: Interactive demonstration of a music composition AI.” HCI - Paper | Demo (1st Author)
2019 ISMIR Late-Breaking: “Learning to generate jazz and pop piano music from audio via MIR techniques.” HCI - Paper (2nd Author)
GitHub Organization
Yating Music, Taiwan AI Labs
2019 April - 2024 December
[GitHub]
From 2019, as one of the core members of a research team at Taiwan, we open-sourced AI technologies for music before the dawn of the generative AI era.
Toolkits
miditoolkit (254 stars) - GitHub toolkit - A python toolkit for handling MIDI I/O in ticks, the native time unit of the MIDI protocol.
Lead Sheet Dataset (119 stars) - GitHub dataset -A web crawler code for lead sheets (melody&chord) from Hooktheory.
ReaRender (98 stars) - GitHub toolkit - A python toolkit for automatic audio/MIDI rendering using REAPER.
Technical staff responsible for deploying text-to-video and text-to-music models in the exhibition. Located in Hualien, a beautiful town surrounded by mountains and sea.
Research & Work Experience Overview
2024 December
[PDF]
This presentation provides an overview of my research and professional experience up to 2024, along with my technical stack and the key industry trends over time.
My Research Notes on Audio Effect Modeling
2022 December
[PDF] | [Paper]
These slides document my experience in digital signal processing, audio effect modeling, and the applications of neural network in these domains.
