I am an assistant professor in the Computer Science Department at Carnegie Mellon University. I am also a part-time research scientist at Google Magenta.

My research goal is to build music technology that enables a broader set of users to engage with music on a deeper level. To this end, I improve ML methods for generative modeling of music and other multimedia, and build interactive systems which allow anyone to leverage resultant models.

Previously, I was a postdoc at Stanford CS advised by Percy Liang. Before that, I completed a PhD at UCSD under Miller Puckette and Julian McAuley.

Selected Publications (Full List)

• Chris Donahue, John Thickstun, Percy Liang Melody Transcription via Generative Pre-training In ISMIR, 2022. [pdf, arXiv, sound examples, code and data]
• Karan Goel, Albert Gu, Chris Donahue, Christopher Ré It's Raw! Audio Generation with State-Space Models In ICML (Long Talk; Top 2%), 2022. [pdf, arXiv, sound examples, code]
• Rodrigo Castellon*, Chris Donahue*, Percy Liang Codified Audio Language Modeling Learns Useful Representations for Music Information Retrieval In ISMIR (Best Paper Runner-up), 2021. [pdf, arXiv, code]
• Hao-Wen Dong, Chris Donahue, Taylor Berg-Kirkpatrick, Julian McAuley Towards Automatic Instrumentation by Learning to Separate Parts in Symbolic Multitrack Music In ISMIR, 2021. [pdf, arXiv, sound examples, code]
• Mina Lee*, Chris Donahue*, Robin Jia, Alexander Iyabor, Percy Liang Swords: A Benchmark for Lexical Substitution with Improved Data Coverage and Quality In NAACL, 2021. [pdf, arXiv, code]
• Chris Donahue, Mina Lee, Percy Liang Enabling Language Models to Fill in the Blanks In ACL, 2020. [pdf, arXiv, demo, code]
• Chris Donahue, Huanru Henry Mao, Yiting Ethan Li, Garrison W. Cottrell, Julian McAuley LakhNES: Improving Multi-instrumental Music Generation with Cross-domain Pre-training In ISMIR, 2019. [pdf, arXiv, code, music examples]
• Paarth Neekhara*, Chris Donahue*, Miller Puckette, Shlomo Dubnov, Julian McAuley Expediting TTS Synthesis with Adversarial Vocoding In INTERSPEECH (Oral), 2019. [pdf, arXiv, code, sound examples]
• Chris Donahue, Ian Simon, Sander Dieleman Piano Genie In ACM IUI, 2019. [pdf, arXiv, blog post, videos, demo, code]
• Chris Donahue, Julian McAuley, Miller Puckette Adversarial Audio Synthesis In ICLR, 2019. [pdf, arXiv, code, demo, sound examples, notebook]
• Jesse Engel, Kumar Krishna Agrawal, Shuo Chen, Ishaan Gulrajani, Chris Donahue, Adam Roberts GANSynth: Adversarial Neural Audio Synthesis In ICLR, 2019. [pdf, blog post, code, sound examples]
• Chris Donahue, Huanru Henry Mao, Julian McAuley The NES Music Database: A Multi-instrumental Dataset with Expressive Performance Attributes In ISMIR, 2018. [pdf, arXiv, dataset, code, notebook]
• Chris Donahue, Zachary C. Lipton, Akshay Balsubramani, Julian McAuley Semantically Decomposing the Latent Spaces of Generative Adversarial Networks In ICLR, 2018. [pdf, arXiv, code, demo]
• Chris Donahue, Bo Li, Rohit Prabhavalkar Exploring Speech Enhancement with Generative Adversarial Networks for Robust Speech Recognition In ICASSP (Oral), 2018. [pdf, arXiv]
• Chris Donahue, Zachary C. Lipton, Julian McAuley Dance Dance Convolution In ICML, 2017. [pdf, arXiv, dataset, code, demo]
* Indicates equal contribution

Work Experience

• (2022-Current) Research Scientist at Google Magenta.
• (Summer 2018) Internship at Google Magenta (Music generation w/ Ian Simon and Sander Dieleman)
• (Summer 2017) Internship at Google (Speech recognition w/ Bo Li and Rohit Prabhavalkar)
• (Summer 2016) Internship at Google Search
• (Summer 2015) Internship at Google Play Music (MIR w/ Nicolas Boulanger-Lewandowski)
• (2011-2014) Mentor for UT Freshman Research Initiative w/ Joel Lehman and Risto Miikkulainen
• (Summers 2011-2014) Internships at UT Applied Research Laboratories, Qualcomm, and two startups

Media Coverage

• UploadVR Beat Sage Update Adds 90 Degree Levels, Walls And Single Saber Mode
• UploadVR Get Rhythm: How Beat Sage Uses AI To Create Beat Saber Maps
• Road to VR This 'Beat Saber' Project Uses AI to Generate Custom Beat Maps for Any Song
• UploadVR New AI Tool Turns Any Song Into A Custom Beat Saber Map, And It Really Works

---

• Stereogum Watch The Flaming Lips Play A Bowl Of Fruit At Google I/O

---

• Business Insider A Google intern helped build an AI tool inspired by 'Guitar Hero' to let rookies play piano
• The Verge Google’s AI-powered Piano Genie lets anyone improvise perfectly by bashing buttons
• Evening Standard Piano Genie: Google's AI programme is like Guitar Hero for the piano world
• Engadget Google’s Piano Genie lets anyone improvise classical music

---

• MIT Tech Review Machine-Learning Algorithm Watches DDR, Then Creates Dances of Its Own
• The Verge Scientists have taught a neural network to choreograph Dance Dance Revolution levels
• The Register Yet another job menaced by AI! Uh, wait, it says here... Dance Dance Revolution designers
• Vice This Machine Learned to Choreograph by Watching Dance Dance Revolution

Other

• (2020) Released Beat Sage, a web service for automatically creating Beat Saber levels (link)
• (2020) Download files from Google Drive on the command line (link, code)
• (2019) PhD dissertation on music, AI, and interaction (pdf)
• (2018) Transcribe a batch of solo piano recordings to MIDI (link)
• (2017) PhD qualifying examination (pdf)
• (2016) Master's thesis on convolution-based cross-synthesis, (pdf, code, vst)
• (2015) Prototype for MOOC on computer music fundamentals using Web Audio API (link)
• (2015) Mobile-friendly, networked musical controller (demo)
• (2014) Multichannel convolution reverb plugin (screenshot, code, windows vst)
• (2013) Undergraduate thesis on musical instrument mimicry (pdf, examples, vst)
• (2013) OpenGL 3D spectrogram (page, code)
• (2012-2014) Played keyboard for Food Group

Last updated 2023/8/28