SemanticAudio: Audio Generation and Editing in Semantic Space

TL;DR

SemanticAudio introduces a high-level semantic space for audio generation and editing, improving alignment with textual descriptions and enabling precise, training-free attribute modifications through a novel two-stage architecture.

Contribution

It proposes a new semantic space and a two-stage Flow Matching architecture for improved text-to-audio generation and editing, with a training-free editing mechanism.

Findings

Outperforms existing methods in semantic alignment

Enables precise attribute-level audio editing without retraining

Demonstrates high-fidelity audio generation from semantic sketches

Abstract

In recent years, Text-to-Audio Generation has achieved remarkable progress, offering sound creators powerful tools to transform textual inspirations into vivid audio. However, existing models predominantly operate directly in the acoustic latent space of a Variational Autoencoder (VAE), often leading to suboptimal alignment between generated audio and textual descriptions. In this paper, we introduce SemanticAudio, a novel framework that conducts both audio generation and editing directly in a high-level semantic space. We define this semantic space as a compact representation capturing the global identity and temporal sequence of sound events, distinct from fine-grained acoustic details. SemanticAudio employs a two-stage Flow Matching architecture: the Semantic Planner first generates these compact semantic features to sketch the global semantic layout, and the Acoustic Synthesizer subsequently produces high-fidelity acoustic latents conditioned on this semantic plan. Leveraging this decoupled design, we further introduce a training-free text-guided editing mechanism that enables precise attribute-level modifications on general audio without retraining. Specifically, this is achieved by steering the semantic generation trajectory via the difference of velocity fields derived from source and target text prompts. Extensive experiments demonstrate that SemanticAudio surpasses existing mainstream approaches in semantic alignment. Demo available at: https://semanticaudio1.github.io/