Whisper Large Galician is an automatic speech recognition (ASR) model for Galician (gl) speech. It is fine-tuned from [openai/whisper-large] on the Galician portion of Mozilla Common Voice 13.0, achieving a Word Error Rate (WER) of 6.94% on the Common Voice evaluation split.

This model provides high-accuracy transcription for large-scale Galician ASR applications.

Model description

Architecture: Transformer-based encoder–decoder (Whisper)
Base model: openai/whisper-large
Language: Galician (gl)
Task: Automatic Speech Recognition (ASR)
Output: Text transcription in Galician
Decoding: Autoregressive sequence-to-sequence decoding

The large model leverages Whisper’s multilingual pretraining and is fine-tuned on Galician speech data to deliver high-quality transcription suitable for research, media, and accessibility applications.

Intended use

Primary use cases

High-accuracy transcription of Galician audio recordings
Offline or batch ASR pipelines
Research and development in Galician ASR
Media, educational, and archival transcription tasks

Intended users

Researchers working on Galician or low-resource ASR
Developers building Galician speech applications
Academic or institutional users

Out-of-scope use

Real-time or low-latency ASR without optimization
Speech translation tasks
Safety-critical applications without validation

Limitations and known issues

Performance may degrade on:
- Noisy or low-quality recordings
- Conversational or spontaneous speech
- Accents underrepresented in Common Voice
Transcription errors may still occur under challenging acoustic conditions
Dataset biases from Common Voice may be reflected in outputs

Users are encouraged to evaluate the model on their own data before deployment.

Training and evaluation data

Training data

Dataset: Mozilla Common Voice 13.0 (Galician subset)
Data type: Crowd-sourced, read speech
Preprocessing:
- Audio resampled to 16 kHz
- Text normalized using Whisper tokenizer
- Filtering of invalid or problematic samples

Evaluation data

Dataset: Mozilla Common Voice 13.0 (Galician evaluation split)
Metric: Word Error Rate (WER)

Evaluation results

Metric	Value
WER (eval)	6.94%

This reflects the expected performance of a large Whisper model fine-tuned for Galician.

Training procedure

Training hyperparameters

Learning rate: 1e-5
Optimizer: Adam (β1=0.9, β2=0.999, ε=1e-8)
LR scheduler: Linear
Warmup steps: 500
Training steps: 20,000
Train batch size: 32
Evaluation batch size: 16
Gradient accumulation steps: 2
Total train batch size: 64
Seed: 42

Training results (summary)

Training Loss	Epoch	Step	Validation Loss	WER
0.0126	4.01	1000	0.2128	8.3558
0.0032	9.01	2000	0.2262	6.9416
0.0022	14.01	3000	0.2528	7.1123
0.0025	19.01	4000	0.2643	7.3641
0.0015	24.01	5000	0.2596	7.3365
0.0014	29.01	6000	0.2723	7.6366
0.0008	34.01	7000	0.2778	7.6090
0.0003	39.01	8000	0.2880	7.2261
0.0004	44.01	9000	0.2920	7.6745
0.0001	49.01	10000	0.2854	7.4089
0.0	54.01	11000	0.3027	7.4365
0.0	59.01	12000	0.3159	7.4055
0.0	64.01	13000	0.3242	7.3693
0.0	69.01	14000	0.3312	7.3072
0.0	74.01	15000	0.3379	7.0226
0.0	79.01	16000	0.3442	7.0019
0.0	84.01	17000	0.3500	6.9933
0.0	89.01	18000	0.3550	6.9605
0.0	94.01	19000	0.3589	6.9467
0.0	99.01	20000	0.3605	6.9398

Framework versions

Transformers 4.33.0.dev0
PyTorch 2.0.1+cu117
Datasets 2.14.4
Tokenizers 0.13.3

How to use

python

from transformers import pipeline

hf_model = "HiTZ/whisper-large-gl"  # replace with actual repo ID
device = 0  # set to -1 for CPU

pipe = pipeline(
    task="automatic-speech-recognition",
    model=hf_model,
    device=device
)

result = pipe("audio.wav")
print(result["text"])

Ethical considerations and risks

This model transcribes speech and may process personal data.
Users should ensure compliance with applicable data protection laws (e.g., GDPR).
The model should not be used for surveillance or non-consensual audio processing.

Citation

If you use this model in your research, please cite:

bibtex

@misc{dezuazo2025whisperlmimprovingasrmodels,
  title={Whisper-LM: Improving ASR Models with Language Models for Low-Resource Languages},
  author={Xabier de Zuazo and Eva Navas and Ibon Saratxaga and Inma Hernáez Rioja},
  year={2025},
  eprint={2503.23542},
  archivePrefix={arXiv},
  primaryClass={cs.CL}
}

Please, check the related paper preprint in arXiv:2503.23542 for more details.

License

This model is available under the Apache-2.0 License. You are free to use, modify, and distribute this model as long as you credit the original creators.

Contact and attribution

Fine-tuning and evaluation: HiTZ/Aholab - Basque Center for Language Technology
Base model: OpenAI Whisper
Dataset: Mozilla Common Voice

For questions or issues, please open an issue in the model repository.