pymmdrza/Common-Voice-Speech-26.0-Persian-Clean

Persian Common Voice Clean Dataset

This dataset is a cleaned and prepared subset of the Persian (فارسی - fa) portion of Mozilla Common Voice Scripted Speech, based on cv-corpus-26.0-2026-06-12.

The cleaned release contains 34,134 audio clips, representing approximately 43.105 hours of speech, equal to 2,586.303 minutes. The clips are associated with approximately 34,134 validated Persian sentences and come from 3,791 speakers.

The original Persian Common Voice release contains approximately 373.24 validated hours. This repository provides a filtered, cleaned, and training-ready subset suitable for speech processing, ASR, TTS, and Persian language research.

Dataset Summary

Field	Value
Language	Persian / فارسی
Language code	fa
Source corpus	Mozilla Common Voice Scripted Speech
Source version	cv-corpus-26.0-2026-06-12
Cleaned clips	34,134
Cleaned duration	43.105 hours
Cleaned duration minutes	2,586.303 minutes
Cleaned duration HH:MM:SS	43:06:18
Speakers	3,791
Approx. text-audio pairs	34,134
License	cc0-1.0

Intended Uses

This dataset is intended for:

• Persian automatic speech recognition research
• Persian text-to-speech research
• Speech preprocessing experiments
• Speaker and acoustic analysis
• Persian/Farsi language technology research

Dataset Structure

This repository keeps the cleaned dataset files in the same prepared structure used locally for F5-TTS-style training and audio processing.

Typical contents may include:

ext audio files metadata or manifest files reports/ speakers.csv dataset summary files

If you want to load this dataset with Hugging Face datasets, inspect the available metadata files in the repository and map the audio path column to the transcript column.

Source and Processing

The source data is derived from Mozilla Common Voice Scripted Speech for Persian.

Cleaning and preparation may include:

• keeping validated Persian speech samples
• filtering unusable or missing audio
• preparing metadata for speech training
• computing speaker-level duration statistics
• organizing files for downstream ASR/TTS workflows

Limitations

This is a cleaned subset, not the full Persian Common Voice release. It may not preserve the full demographic, acoustic, or dialectal distribution of the original corpus. Users should evaluate suitability before training production models.

License and Attribution

This dataset is derived from Mozilla Common Voice, which is released under the CC0-1.0 public domain dedication. Please also review Mozilla Common Voice licensing and attribution guidance before redistributing or using this dataset in production.

Citation

If you use this dataset, please cite Mozilla Common Voice and this cleaned dataset repository.

@dataset{persian_common_voice_clean,
  title  = {Persian Common Voice Clean},
  author = {Mmdrza},
  year   = {2026},
  url    = {https://huggingface.co/datasets/pymmdrza/Common-Voice-Speech-26.0-Persian-Clean},
  note   = {Cleaned Persian subset derived from Mozilla Common Voice Scripted Speech cv-corpus-26.0-2026-06-12}
}