---
license: apache-2.0
base_model: openai/whisper-large-v3
datasets:
- burkimbia/moore-speech-dataset-public
language:
- mos
tags:
- automatic-speech-recognition
- asr
- moore
- burkina-faso
- african-languages
pipeline_tag: automatic-speech-recognition
---


# BIA-WHISPER-LARGE-SACHI_V2

## Model Description

BIA-WHISPER-LARGE-SACHI_V2 is an automatic speech recognition (ASR) model for the Moore language, a Gur language spoken in Burkina Faso. This model is based on OpenAI's Whisper architecture and fine-tuned on high-quality Moore speech data.

## Moore Language
- **Language**: Moore (Mooré)
- **Language Code**: `mos`
- **Region**: Burkina Faso, West Africa
- **Speakers**: ~5 million native speakers

## Inference

You can use this model for Moore speech-to-text transcription as follows:

```python
from transformers import pipeline

asr = pipeline("automatic-speech-recognition", model="BIA-WHISPER-LARGE-SACHI_V2")
result = asr("path/to/audio.wav")
print(result["text"])
```



## Training Data
- **Dataset**: Custom Moore speech dataset
- **Hours**: ~100 hours of high-quality Moore audio
- **Sources**: Moore/Burkinabé Bible (JW.org, Bible.com), community contributions

## Model Details
- **Architecture**: Whisper v3 (OpenAI)
- **Fine-tuning**: QLORA

## Citation

If you use this model, please cite:

```bibtex
@misc{bia-whisper-large-sachi-v2,
	title={BIA-WHISPER-LARGE-SACHI_V1},
	author={Salif SAWADOGO at Burkimbia},
	year={2025},
	howpublished={\url{https://huggingface.co/your-model-repo}}
}
```

## Acknowledgments
- [Deezer Spleeter](https://github.com/deezer/spleeter) and other Deezer tools for audio processing
- Moore/Burkinabé Bible audio from [JW.org](https://www.jw.org/) and [Bible.com](https://www.bible.com/)
- The Moore language community for providing around 100 hours of audio data