HiTZ/whisper-tiny-eu

Whisper Tiny Basque

Model summary

Whisper Tiny Basque is an automatic speech recognition (ASR) model for Basque (eu) speech. It is fine-tuned from [openai/whisper-tiny] on the Basque portion of Mozilla Common Voice 13.0, achieving a Word Error Rate (WER) of 32.27% on the Common Voice evaluation split.

The model is designed for lightweight transcription of Basque speech, prioritizing low computational cost over transcription accuracy.

---

Model description

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

This model leverages Whisper’s multilingual pretraining and is further fine-tuned on Basque speech data to enable ASR for a low-resource language, using a compact model size suitable for constrained environments.

---

Intended use

Primary use cases

• Basque speech transcription in low-resource or experimental settings
• Lightweight ASR pipelines with limited computational resources
• Research on Basque ASR and low-resource speech recognition
• Dataset exploration and preprocessing

Intended users

• Researchers working on Basque or low-resource ASR
• Developers experimenting with compact ASR models
• Academic and educational use

Out-of-scope use

• High-accuracy transcription requirements
• Real-time or production-grade ASR without further optimization
• Speech translation tasks
• Safety-critical applications

---

Limitations and known issues

• Relatively high WER compared to larger Whisper variants
• Performance may degrade significantly on:
  • Noisy audio
  • Conversational or spontaneous speech
  • Accents underrepresented in Common Voice
• As a tiny model, it may:
  • Miss words
  • Produce incomplete or inaccurate transcriptions
• 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 (Basque subset)
• Data type: Crowd-sourced, read speech
• Preprocessing:
  • Audio resampled to 16 kHz
  • Text normalized using Whisper tokenizer
  • Filtering of invalid samples

Evaluation data

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

---

Evaluation results

Metric	Value
WER (eval)	32.27%

These results reflect the expected performance of a tiny Whisper model fine-tuned for a low-resource language.

---

Training procedure

Training hyperparameters

• Learning rate: 3.75e-5
• Optimizer: Adam (β1=0.9, β2=0.999, ε=1e-8)
• LR scheduler: Linear
• Warmup steps: 500
• Training steps: 5,000
• Train batch size: 256
• Evaluation batch size: 128
• Seed: 42

Training results (summary)

Training Loss	Epoch	Step	Validation Loss	WER
0.0096	19.0	1000	0.5796	32.8952
0.0011	38.0	2000	0.6522	32.2694
0.0005	57.01	3000	0.6949	33.1403
0.0003	76.01	4000	0.7217	33.0734
0.0003	96.0	5000	0.7321	33.1585

---

Framework versions

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

---

How to use

from transformers import pipeline

hf_model = "HiTZ/whisper-tiny-eu"  # 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:

@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.

Funding

This project with reference 2022/TL22/00215335 has been parcially funded by the Ministerio de Transformación Digital and by the Plan de Recuperación, Transformación y Resiliencia – Funded by the European Union – NextGenerationEU ILENIA and by the project IkerGaitu funded by the Basque Government. This model was trained at Hyperion, one of the high-performance computing (HPC) systems hosted by the DIPC Supercomputing Center.