---
language:
- nl
---

# Model

This repository contains the first version of our Automatic Speech Recognition and Subtitle Generation model, trained on 2000 hours of Flemish broadcast subtitled speech data. 
It can generate both a verbatim and a subtitle transcription.
It outputs both modalities with separate decoders.

**Version**: September 2023

# Usage

This repository only hosts the pre-trained model itself and the configuration files. 
Using this model requires installation of our Flemish Speech&Subtitling branch of the ESPnet toolkit, which you can find on Github [**here**](https://github.com/nelfproject/ASR_espnet_v1).
The Github page will contain all further details to use the model. To download the model, see the instructions [here](https://huggingface.co/docs/hub/models-downloading).

# Webservice

This model can also be accessed through the [webservice of the NeLF Project](https://nelfproject.be/web_service.php). After requesting access, you can upload audio or video files and they will be transcribed according to the desired settings. 

# Citation

If you use this model, please cite the research paper: (Will be added shortly).

# Contact

Jakob Poncelet: jakob.poncelet@kuleuven.be