dvtk 0.1.0

Lightweight package to validate/measure audio files in local or S3 directories.

Usage

1. Install

pip install dvtk

2. Run

Run dvtk to validate audio files in local directories.

dvtk \
  --paths /path/to/directory/ \
  --ext .wav \
  --metrics dvtk.default_metrics:sr,duration,max_amplitude,dnsmos_speech \
  --output dv_result.csv \
  --parallel \
  --max-workers 4

Run dvtk to validate audio files in S3 directories.

dvtk \
  --paths s3://bucket-name/path/to/directory/ \
  --ext .wav \
  --metrics dvtk.default_metrics:sr,duration,max_amplitude,dnsmos_speech \
  --output dv_result.csv \
  --parallel \
  --max-workers 4

Remarks

• You can specify multiple paths and extensions (e.g., --ext .wav .flac).
• You can specify the AWS region to use (--region-name).
• You can specify the AWS profile to use (--aws-profile).
• More default metrics are available in dvtk.default_metrics (e.g., reference-free PESQ, reference-free SI-SDR for speech).

Using Custom Metrics

1. Create a small python module (e.g., mymetrics.py) in your working directory:

# mymetrics.py
import numpy as np

def custom_metric(audio: np.ndarray, sr: int) -> float:
    ...  # compute the custom metric
    return metric  # float

2. Run the CLI:

dvtk \
  --paths /path/to/directory/ \
  --ext .wav \
  --metrics dvtk.default_metrics:duration,max_amplitude,dnsmos_speech \
            mymetrics:custom_metric \
  --output dv_result.csv \
  --parallel \
  --max-workers 4

TODOs

• [] Support more storage backends (e.g., Google Cloud Storage)
• [] On-disk caching or cehckpointing
• [] Progress bar