DeepAFx-ST: Automatic control of audio effects for transferring the production style from one recording to another using differentiable signal processing. arXiv: arxiv.org/abs/2207.08759 web: csteinmetz1.github.io/DeepAFx-ST GitHub: github.com/adobe-research… Nicholas J. Bryan QMUL_IntelSoundEng

Very excited to be helping with this! We have two paper tracks, and an exciting speaker lineup. More info: sites.google.com/view/hscma2024

Two big announcements! (a) We appreciate many inquiries about HSCMA 2024. We are happy to announce that the paper submission deadline is extended to Jan. 10, 2024. Please refer to our CfP for more information: sites.google.com/view/hscma2024… (more below👇)

``DITTO-2: Distilled Diffusion Inference-Time T-Optimization for Music Generation,'' Zachary Novack, Julian McAuley, Taylor Berg-Kirkpatrick, Nicholas Bryan, ift.tt/DTChrSb

Presto! Distilling Steps and Layers for Accelerating Music Generation Despite advances in diffusion-based text-to-music (TTM) methods, efficient, high-quality generation remains a challenge. We introduce Presto!, an approach to inference acceleration for score-based diffusion

Accelerate your diffusion model via step & layer distillation. Check it out applied to music! Work lead by the incredible Zachary Novack while at Adobe Research. Thanks to Ge Zhu, Nicholas J. Bryan, McAuley Lab UCSD, and Taylor Berg-Kirkpatrick.

Introducing "DRAGON: Distributional Rewards Optimize Diffusion Generative Models"! 📖: arxiv.org/abs/2504.15217 🎹: ml-dragon.github.io/web/ A new framework for fine-tuning gen models towards a target distribution. By Yatong Bai w/Jonah Casebeer Somayeh Sojoudi Nicholas J. Bryan

Great audio AEs/codecs exist, but when you need structured latents or a tweaked bottleneck for a downstream task (e.g. generation), retraining is expensive & brittle. We Re-Bottleneck👇

Adobe's #GenerateSoundtrack is LIVE today! 🎉 Studio-quality music for storytellers🎵 * Trained on #Licensed data, * Commercially safe, royalty-free, and cleared for any use, & * Exported with #ContentCredentials for transparency and attribution. Get started: