Skip to content

A toolkit to calculate speech audio quality. Not affiliated with the original authors

License

Notifications You must be signed in to change notification settings

fakerybakery/utmos

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

55 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

NOTE: Not affiliated with original authors of UTMOS!

Unofficial UTMOS Python Package

This is an unofficial Python package for UTMOS (UTokyo-SaruLab MOS Prediction System). This repository is based on the original code. The paper is available here.

What is UTMOS?

UTMOS is designed for calculating the mean opinion score (MOS) for a given voice sample. It can be used to calculate audio quality for datasets.

Note

The score is on a scale of 1 to 5. If you'd like a score on 1 to 100, just multiply the score by 20 (score * 20).

Example: new_score = round(score * 100, 2)

Support

This implementation supports CPU, CUDA, and MPS, as well as ROCm if PyTorch is configured properly. This implementation will automatically use the GPU if available.

Installation

pip install utmos

Usage

CLI (Command Line Interface)

utmos audio.wav

Python API

import utmos
model = utmos.Score() # The model will be automatically downloaded and will automatically utilize the GPU if available.
model.calculate_wav_file('audio_file.wav') # -> Float
# or model.calculate_wav(wav, sample_rate)

License

This software is licensed under the MIT license.