---
license: cc-by-nc-4.0
language:
- multilingual
tags:
- forced-alignment
- mms
- audio
---

# Converted MMS Model for Forced Alignment

This repository contains the `.safetensors` version of the multilingual forced alignment model from Meta's MMS project, specifically the one used in `torchaudio.pipelines.MMS_FA`.

## Original Model

This model was originally released as part of the Massively Multilingual Speech (MMS) project. All credits go to the original authors.

- **Paper:** [Scaling Speech Technology to 1,000+ Languages](https://arxiv.org/abs/2305.13516)
- **Official Repository:** [facebookresearch/fairseq (MMS examples)](https://github.com/facebookresearch/fairseq/tree/main/examples/mms)
- **Original License:** CC-BY-NC-4.0. This converted version is also released under the same license.

## Conversion Details

The original model was provided in `.pt` format. It was loaded using `torchaudio.pipelines.MMS_FA.get_model(with_star=False)` and then saved to `.safetensors` format.