AMBIT brings unified generative editing to first-order ambisonics
Kam Man Wu and six coauthors submitted a paper introducing AMBIT, a framework for native First-Order Ambisonics generation and editing. It combines executable ScenePlans, a shared Transfusion backbone, an FOA-aware continuous VAE, and on-policy self-distillation for planning and rendering. The authors report better content, spatial quality, acoustic fidelity, and preservation of unedited sources than compared baselines across speech, music, and sound.
AMBIT offers a unified approach to planning, generating, and editing spatial audio while preserving unedited elements, but its reported advantages currently come only from the paper’s abstract and author comparisons.
Computer Science > Sound
arXiv:2610.34415 (cs)
[Submitted on 25 Oct 2026]
Title:Spatial Audio Generation and Editing with Executable Plans and Self-Distilling Transfusion
Authors:Kam Man Wu, Yu Zhang, Pengjun Fang, Mingfei Chen, Weijia Chen, Zeyue Tian, Qifeng Chen
View a PDF of the paper titled Spatial Audio Generation and Editing with Executable Plans and Self-Distilling Transfusion, by Kam Man Wu and 6 other authors
#AMBIT#Transfusion#ScenePlans#FOA-aware continuous VAE#On-Policy Self-Distillation