Help Center/ ModelArts/ Data Preparation/ Dataset Format Requirements/ Format Requirements for Audio Datasets
Updated on 2026-07-03 GMT+08:00
View PDF
Share

Format Requirements for Audio Datasets

ModelArts supports the creation of audio datasets. Table 1 lists the format requirements.

Constraints

  • Import from OBS: The size of a single file or compressed package cannot exceed 20 GB. If multiple files are imported, the total file size cannot exceed 20 GB.
  • Local import: The size of a single file cannot exceed 1 GB, and the number of files cannot exceed 20.
  • JSONL files support only UTF-8 encoding.

Table 1 Format Requirements for Audio Datasets

File Content

File Format

Requirement

Audio

Audio + JSONL (optional)
  • Audio file: The .mp3, .flac, .wav, .opus, .aac and .m4a files are supported. Audio files can be stored in the root directory or a lower-level directory.

    Annotation file format: Optional. UTF-8-encoded JSONL files are supported. Each line describes the relative path of an audio file in the dataset and other information.

The following is an example.

For details about the annotation file in JSONL format, refer to the following:

{"audio_name":"dir/001.mp3","caption":"1"}
{"audio_name":"dir/002.flac","caption":"2"}
{"audio_name":"dir/003.wav","caption":"3"}
Parent Topic: Dataset Format Requirements