← Back to all news
arXiv AI

PhysWave: Physics-Guided Latent Diffusion Models for Controllable Spatial Audio Generation

PhysWave is a physics-guided latent diffusion model designed for controllable text-to-First-Order Ambisonics (FOA) audio generation. It integrates natural-language and trajectory control via a shared waypoint-caption representation and incorporates two differentiable acoustic priors—spherical-harmonic direction consistency and inverse-square distance consistency—into diffusion training. The authors also introduce a 300K-clip FOA dataset and demonstrate that these priors improve spatial consistency while preserving audio quality, with potential use as inference-time guidance for training-free refinement.

Related stories

arXiv AI
Sep 7

Pitch-class Steering for Diffusion-based Music Generation via Latent-space Probes

The paper introduces a lightweight technique to steer the pitch content of audio generated by the Stable Audio Open diffusion model. A small convolutional probe (~125k parameters) is trained to decode frame‑level pitch‑class activations from the model’s latent space using paired audio and MIDI data. During inference, the frozen probe acts as a differentiable loss, guiding generation toward a user‑specified pitch‑class sequence without retraining the base model, and improves melodic coherence by 2.4× over the unguided baseline.

By Yushi Ye, Wilson Zheng, Yongyi Zang
diffusion
More like this →
arXiv AI
Sep 7

SCAPES: Semantically Conditioned Autoregressive Prior for Environmental Sounds

SCAPES is a lightweight, resource‑efficient generative model that synthesizes high‑fidelity environmental sounds with high‑level semantic control. It operates on the continuous latent manifold of a neural audio codec, using a segmentation strategy and a Continuous Normalizing Flow to model latent trajectories. A 36‑million‑parameter instance can be trained on limited, uncurated data with a single consumer‑grade GPU, achieving convergence in roughly twice the source audio duration and enabling smooth semantic interpolation.

By Esteban Guti\'errez, Lonce Wyse, Frederic Font, Xavier Serra
diffusioncomputer-visionnlp
More like this →
arXiv Machine Learning
Aug 19

From Diffusion to Flow: Efficient Motion Generation in MotionGPT3

The paper compares diffusion and rectified flow objectives within the MotionGPT3 framework for text-driven motion generation. Experiments on HumanML3D show that rectified flow converges faster, achieves strong test performance earlier, and matches or exceeds diffusion quality while requiring fewer sampling steps. The study isolates the generative objective’s impact, demonstrating that rectified flow’s benefits transfer to continuous-latent motion generation.

By Jaymin Bhan, JiHong Jeon, SangYeop Jeong
diffusion
More like this →
arXiv Machine Learning
Sep 10

BinauralVAE: Spatial Audio Reconstruction For World Models

BinauralVAE is an open‑source pipeline that reconstructs spatial audio using various Variational Autoencoder architectures, including complex‑valued variants, to learn latent representations of binaural signals. The project builds on realistic acoustic data from a simulated robot navigating an environment, providing a foundation for audio‑centric world models. It aims to map the causal link between navigational actions and their acoustic outcomes, positioning sound as a complementary modality for spatial awareness.

By Luis Vitor Zerkowski, Luiz Velho
robotics
More like this →