Whisper Small Spanish is an automatic speech recognition (ASR) model for Spanish (es), fine-tuned from [openai/whisper-small] on the Spanish subset of Mozilla Common Voice 13.0. It achieves a Word Error Rate (WER) of 8.2668% on the evaluation split.

This model provides a good balance between transcription accuracy and computational efficiency, suitable for applications requiring relatively low-latency ASR with decent quality.

Model description

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

Compared to Whisper Base, this model is slightly larger and generally more accurate, particularly for standard read Spanish.

Intended use

Primary use cases

Real-time or batch transcription of Spanish speech
Research or experimentation with Spanish ASR
Applications with moderate hardware resources where Whisper Medium or Large is too heavy

Limitations

Performance may degrade for:
- Noisy or overlapping speech
- Regional accents or dialects not well represented in Common Voice
- Very fast conversational speech
Not recommended for safety-critical or professional-level transcription tasks.

Training and evaluation data

Dataset: Mozilla Common Voice 13.0 (Spanish subset)
Data type: Crowd-sourced read speech
Preprocessing:
- Audio resampled to 16 kHz
- Text normalized using Whisper tokenizer
- Invalid or corrupted samples removed
Evaluation metric: Word Error Rate (WER) on held-out evaluation set

Evaluation results

Metric	Value
WER (eval)	8.2668%

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: 5000
Train batch size: 64
Eval batch size: 32
Seed: 42

Training results (summary)

Training Loss	Epoch	Step	Validation Loss	WER
0.1320	2.0	1000	0.2461	9.5267
0.1288	4.01	2000	0.2251	8.5215
0.0814	6.01	3000	0.2212	8.2668
0.0905	8.01	4000	0.2310	8.4997
0.0319	10.02	5000	0.2358	8.5343

Framework versions

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

Example usage

python

from transformers import pipeline

hf_model = "HiTZ/whisper-small-es"  # replace with actual repo ID
device = 0  # -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.