batch_scripts_guide.md

⚙️ Batch Scripts Guide

1. In the scripts for audio reconstruction, encoding, and embedding extraction, you need to provide a JSON file specifying the audio files to be processed.
   • Each entry in the JSON file should contain an index field (used as the basename for storing processed outputs) and a wav_path field (indicating the path to the input audio). An example can be found in example/batch_script_data.jsonl.
2. To explore speech decoupling experiments, we introduce two parameters in the decoding script: acoustic_masked and semantic_masked.
   • You can toggle these options to mask the corresponding encoding streams and observe the reconstruction results.
   • Note that when setting acoustic_masked=true, it is recommended to also set normalize=true to ensure the output volume remains audible.