Overview

Audio editing changes particular qualities or events in recordings—such as words, speaker style, instruments, sound effects, or spatial placement—while aiming to keep unrelated content intact. Common approaches use diffusion or flow models for text-guided editing and inpainting, speech models for precise vocal changes, and latent audio tokens or signal-processing methods for localized control. Unified instruction-following systems increasingly combine audio understanding with generation, often using dialogue, reference audio, or visual context to guide edits. A central challenge is balancing the requested change with preservation of timing, identity, musical structure, and background, while benchmarks assess these goals separately and show that complex, compositional edits remain difficult.

Latest papers

CardsList

62 papers12 papers in the last four weeks, up 20% on the four weeks before. 0.1% of all new papers.

  1. SemanticAudio: Audio Generation and Editing in Semantic Space

    Jan 29, 2026Zheqi Dai, Guangyan Zhang, Haolin He +5Text-to-Audio GenerationAudio Editing

  2. CosyEdit: Unlocking End-to-End Speech Editing Capability from Zero-Shot Text-to-Speech Models

    Jan 8, 2026Junyang Chen, Yuhang Jia, Hui Wang +2Audio EditingSpeech Editing