|
|
--- |
|
|
license: open-mdw |
|
|
language: |
|
|
- lb |
|
|
base_model: |
|
|
- openai/whisper-small |
|
|
pipeline_tag: automatic-speech-recognition |
|
|
--- |
|
|
|
|
|
# unilux/whisper-small-v1-luxembourgish |
|
|
|
|
|
## Model Card |
|
|
|
|
|
### 🧠 Model Details |
|
|
- **Model name:** whisper-small-v1-luxembourgish |
|
|
- **Organization:** University of Luxembourg — Department of Humanities |
|
|
- **Project:** [Luxembourgish Automatic Speech Recognition (LuxASR)](https://luxasr.uni.lu/) |
|
|
- **Type:** Speech-to-Text (ASR) |
|
|
- **Language:** Luxembourgish (`lb`) |
|
|
- **Architecture:** Whisper (Small) |
|
|
- **Model size:** ~244M parameters |
|
|
- **License:** [Open Model, Data & Weights (open-mdw)](https://www.openmdw.org) |
|
|
|
|
|
This model is part of the **LuxASR** open model family for Luxembourgish speech recognition. Fine-tuned on Luxembourgish audio–text pairs (≈150+ hours total initiative scale). |
|
|
|
|
|
The *tiny*, *base*, *small*, and *medium* models are open-sourced; the larger flagship LuxASR model, used in the webservice, the API and the iOS and Android apps, remains closed-source. |
|
|
|
|
|
--- |
|
|
|
|
|
### 🚀 Intended Use |
|
|
- Transcribe Luxembourgish speech into text. |
|
|
- Research and development of Luxembourgish ASR. |
|
|
- Accessibility and media transcription. |
|
|
|
|
|
--- |
|
|
|
|
|
### ⚙️ Usage Example |
|
|
|
|
|
```python |
|
|
from transformers import pipeline |
|
|
|
|
|
pipe = pipeline("automatic-speech-recognition", model="unilux/whisper-small-v1-luxembourgish") |
|
|
result = pipe("example.wav") |
|
|
print(result["text"]) |
|
|
``` |
|
|
|
|
|
### 🧡 Acknowledgements |
|
|
Developed by the **LuxASR** team, University of Luxembourg. |
|
|
See [luxasr.uni.lu](https://luxasr.uni.lu/) for project details. |
