Step-Audio-EditX, the first LLM-level audio editing big data model, has been open-sourced by Step-Star.

Step-Audio-EditX, the first open-source audio editing tool driven by a large language model developed by the Step-Leap Star team, focuses on controlling the emotion, speaking style, and paralinguistic features of audio through iterative methods. Its core technology is trained on large-scale synthetic data, requiring no embedded priors or auxiliary modules, enabling zero-shot text-to-speech functionality.