DegDiT: Controllable Audio Generation with Dynamic Event Graph Guided Diffusion Transformer
By: Yisu Liu , Chenxing Li , Wanqian Zhang and more
Potential Business Impact:
Makes sounds match words exactly, even timing.
Controllable text-to-audio generation aims to synthesize audio from textual descriptions while satisfying user-specified constraints, including event types, temporal sequences, and onset and offset timestamps. This enables precise control over both the content and temporal structure of the generated audio. Despite recent progress, existing methods still face inherent trade-offs among accurate temporal localization, open-vocabulary scalability, and practical efficiency. To address these challenges, we propose DegDiT, a novel dynamic event graph-guided diffusion transformer framework for open-vocabulary controllable audio generation. DegDiT encodes the events in the description as structured dynamic graphs. The nodes in each graph are designed to represent three aspects: semantic features, temporal attributes, and inter-event connections. A graph transformer is employed to integrate these nodes and produce contextualized event embeddings that serve as guidance for the diffusion model. To ensure high-quality and diverse training data, we introduce a quality-balanced data selection pipeline that combines hierarchical event annotation with multi-criteria quality scoring, resulting in a curated dataset with semantic diversity. Furthermore, we present consensus preference optimization, facilitating audio generation through consensus among multiple reward signals. Extensive experiments on AudioCondition, DESED, and AudioTime datasets demonstrate that DegDiT achieves state-of-the-art performances across a variety of objective and subjective evaluation metrics.
Similar Papers
ControlAudio: Tackling Text-Guided, Timing-Indicated and Intelligible Audio Generation via Progressive Diffusion Modeling
Sound
Makes computers talk with perfect timing and clarity.
SSG-Dit: A Spatial Signal Guided Framework for Controllable Video Generation
CV and Pattern Recognition
Makes videos match your exact descriptions.
DiTSE: High-Fidelity Generative Speech Enhancement via Latent Diffusion Transformers
Audio and Speech Processing
Cleans up noisy and echoey voices perfectly.