Implementation:Speechbrain Speechbrain Create Aishell1Mix Metadata

Knowledge Sources	SpeechBrain
Domains	Speech Separation, Data Preparation
Last Updated	2026-02-09 00:00 GMT

Overview

Concrete tool for creating Aishell1Mix metadata files for speech separation provided by the SpeechBrain library.

Description

This script generates metadata CSV files that describe how to mix AISHELL-1 source utterances with WHAM! noise to create the Aishell1Mix speech separation dataset. It reads per-source metadata from both AISHELL-1 and WHAM! directories, randomly pairs speakers, randomizes loudness levels between -33 and -25 LUFS, and writes out CSV metadata files specifying the mixing parameters for each utterance combination. These metadata files are then consumed by the companion create_aishell1mix_from_metadata.py script to generate the actual audio mixtures.

Usage

Use this when generating metadata for the Aishell1Mix dataset prior to creating the actual mixed audio files for speech separation training with SpeechBrain.

Code Reference

Source Location

Repository: SpeechBrain
File: recipes/Aishell1Mix/separation/scripts/create_aishell1mix_metadata.py

Signature

def create_aishell1mix_metadata(
    aishell1_dir, aishell1_md_dir, wham_dir, wham_md_dir, md_dir, n_src
):

CLI Interface

python create_aishell1mix_metadata.py \
    --aishell1_dir /path/to/aishell1/data_aishell/wav \
    --aishell1_md_dir /path/to/metadata/aishell1 \
    --wham_dir /path/to/wham_noise \
    --wham_md_dir /path/to/metadata/wham_noise \
    --metadata_outdir /path/to/aishell1mix/metadata \
    --n_src 2

I/O Contract

Inputs

Name	Type	Required	Description
aishell1_dir	str	Yes	Path to AISHELL-1 root directory containing WAV files
aishell1_md_dir	str	Yes	Path to AISHELL-1 metadata directory
wham_dir	str	Yes	Path to WHAM! noise root directory
wham_md_dir	str	Yes	Path to WHAM! noise metadata directory
metadata_outdir	str	Yes	Where Aishell1Mix metadata files will be stored
n_src	int	No	Number of sources desired to create the mixture (default: 2)

Outputs

Name	Type	Description
Metadata CSV files	CSV Files	Metadata files describing mixing parameters for each utterance combination across train/dev/test splits

Usage Examples

# Typically invoked via command line:
# python create_aishell1mix_metadata.py --aishell1_dir /data/aishell1/wav --wham_dir /data/wham_noise --metadata_outdir /data/aishell1mix/metadata

# Or programmatically:
from create_aishell1mix_metadata import create_aishell1mix_metadata

create_aishell1mix_metadata(
    aishell1_dir="/data/aishell1/data_aishell/wav",
    aishell1_md_dir="/data/metadata/aishell1",
    wham_dir="/data/wham_noise",
    wham_md_dir="/data/metadata/wham_noise",
    md_dir="/data/aishell1mix/metadata",
    n_src=2,
)

Related Pages

Implementation:Speechbrain_Speechbrain_Create_Aishell1Mix_From_Metadata -- Companion script for generating audio from metadata
Principle:Speechbrain_Speechbrain_Dataset_Specific_Data_Preparation

Page Connections

Double-click a node to navigate. Hold to expand connections.

Principle

Implementation

Heuristic

Environment