News

Oct 2026 Smorph, playable sound morphing with diffusion models, will appear at ISMIR 2026, with Annie Chu.

Oct 2026 Released CrossEdit, which uses cross-modal training for rich audio-visual editing.

Sep 2026 TAC, our timestamped audio captioning model, was accepted to NeurIPS 2026!

Aug 2026 We launched Generate Soundscape in Premiere (beta), which generates a picture-synced soundscape for your video as four editable stems, based on our MultiFoley research.

May 2026 Five papers at ICASSP 2026 in Barcelona: Taming Audio VAEs via Target-KL Regularization, Generative Audio Extension and Morphing, Mix2Morph, AudioCards, and PromptSep.

Apr 2026 Two papers at CHI 2026: MoSound (Best Paper Honorable Mention) and SoundStager.

Feb 2026 Released AudioChat, a unified model for audio storytelling, editing, and understanding, and TAC, a timestamped audio captioning model.

Oct 2025 Project Sound Stager was an Adobe MAX Sneak. It builds layered soundscapes for a video from its visuals, pacing, and mood, and lets you refine the mix by chatting with an AI sound designer.

Sep 2025 The Rhythm In Anything (audio-prompted drum generation) at ISMIR 2025, and SILA at WASPAA 2025.

Jul 2025 Adobe launched Generate Sound Effects in Firefly, which lets you make sound effects from a text prompt and your own voice, building on our Sketch2Sound research.

Jul 2025 FLAM, frame-wise language-audio modeling, at ICML 2025.

Jun 2025 MultiFoley, video-guided Foley sound generation with multimodal controls, at CVPR 2025.

Apr 2025 Sketch2Sound (with Hugo) and Code Drift at ICASSP 2025.

Oct 2024 Project Super Sonic was an Adobe MAX Sneak. It generates sound effects for video from text, clicks on objects, and vocal imitations.

Dec 2023 I joined Adobe Research as a Senior Research Scientist, working in the Audio AI Lab.

Jul 2023 Descript launches Regenerate (led by Rithesh Kumar), powered by the Descript Audio Codec!

Jun 2023 Released VampNet, (ISMIR 2023) a really fun (and fast) unconditional music generation model with my intern Hugo!

Jun 2023 I had a kid!

May 2023 Released the Descript Audio Codec (NeurIPS 2023), a powerful neural audio codec that can compress audio 90x with minimal quality loss.

Jan 2022 First intern batch at Descript results in Wav2CLIP (ICASSP) and CARGAN (ICLR)!

Jul 2021 Shipped Studio Sound at Descript.

Feb 2021 Shipped Room Tone at Descript.

Jul 2020 Started as a Research Scientist at Descript.

May 2020 I got married!

Sep 2019 Started as a teaching post-doc at Northwestern, teaching ML.

Sep 2019 I defended my PhD! More here!