ebellob/annotated_catalan_common_voice_v17_cleaned_enhanced

Processed Annotated Catalan Common Voice v17 (CleanUNet + FlashSR)

Dataset Summary

This dataset is a processed and enhanced version of: projecte-aina/annotated_catalan_common_voice_v17. Furthermore, as this is a personal project, we give no guarantees that the audio is completely clean from any artifacts or noise the CleanUNet model could not remove. However, we have personally tested the corpus via the fine-tuning of some SOTA speech models and the results have been satisfactory.

It has been created to support high-quality speech research, particularly for:

• Text-to-Speech (TTS)
• Automatic Speech Recognition (ASR)
• Speech enhancement and robustness studies

due to the lack of big and clean speech datasets in Catalan. This work is a derivation of the original Common Voice 17 dataset.

Find samples in: https://erikupv.github.io/CleanUNet-FlashSR-samples/

Main processing steps

1. Quality filtering Only samples with mean_quality > 4.0 are kept. This removes clips with consistently poor subjective or automatic quality ratings.

2. Speech denoising (CleanUNet) All audio samples are processed using CleanUNet, a neural speech enhancement model, to reduce background noise and artifacts while preserving speech content.

3. Audio super-resolution (FlashSR) Enhanced audio is further processed using FlashSR to improve temporal and spectral resolution, resulting in cleaner and more detailed waveforms suitable for modern TTS and ASR pipelines.

The metadata structure and annotations are preserved from the original dataset unless stated otherwise.

---

Dataset Structure

Configurations

This dataset currently provides a single train split.

---

Dataset Features

• client_id: Speaker identifier
• path: Relative audio file path
• sentence_id: Sentence identifier
• sentence: Spoken sentence text
• sentence_domain: Domain/category of the sentence
• up_votes / down_votes: Validation votes
• age: Speaker age group
• gender: Self-reported gender
• accents: Self-reported accent
• variant / variant_norm: Language variant
• locale: Locale identifier
• segment: Segment information
• mean_quality / stdev_quality: Quality statistics (float)
• annotated_*: Annotator-provided metadata
• propagated_*: Automatically propagated metadata
• assigned_accent / assigned_gender: Final assigned labels
• quality: Numeric quality score
• audio: Enhanced audio waveform

---

Dataset Size

• Number of examples: 120,220
• Download size: ~53.3 GB

---

Language

• Catalan (ca)

---

Intended Uses

This dataset is suitable for:

• High-quality Catalan TTS training
• ASR model training and evaluation
• Speech enhancement and super-resolution research
• Accent- and gender-aware speech modeling
• Robustness studies on enhanced speech data

---

Out-of-Scope Uses

• Speaker identification or biometric profiling
• Any attempt to deanonymize speakers
• Uses violating the original Common Voice license or contributor consent

---

Processing Notes

• Audio waveforms are not bit-identical to the original Common Voice release.
• Speech content and textual annotations are unchanged.
• Processing may introduce minor spectral or temporal artifacts inherent to neural enhancement models.

---

License

This dataset inherits the license of the original dataset:

• Creative Commons Attribution 4.0 (CC-BY-4.0)

Please refer to the original dataset page for exact license terms and attribution requirements.

---

Citation

If you use this dataset, please cite both the original dataset and this processed version.

Original dataset

@dataset{aina_annotated_catalan_common_voice_v17,
  title = {Annotated Catalan Common Voice v17},
  author = {Projecte AINA},
  year = {2024},
  url = {https://huggingface.co/datasets/projecte-aina/annotated_catalan_common_voice_v17},
  doi = {10.57967/hf/5679}
}

This dataset

@dataset{processed_ca_common_voice_v17,
  title = {Processed Annotated Catalan Common Voice v17 (CleanUNet + FlashSR)},
  author = {Erik Beltrán},
  year = {2026},
  note = {Enhanced with CleanUNet denoising and FlashSR super-resolution}
}

Tools used

• CleanUNet — Kong et al., "Speech Denoising in the Waveform Domain with Self-Attention", ICASSP 2022.
  Paper: arXiv:2202.07790 · Code: NVIDIA/CleanUNet

• FlashSR — Im & Nam, "FlashSR: One-step Versatile Audio Super-resolution via Diffusion Distillation", 2025.
  Paper: arXiv:2501.10807 · Code: ysharma3501/FlashSR

---

Acknowledgements

• Mozilla Common Voice contributors
• Projecte AINA
• CleanUNet authors (Kong, Ping, Dantrey & Catanzaro)
• FlashSR authors (Jaekwon Im & Juhan Nam)