Loading datasets...
Release Date: 10/20/2026
Format: FLAC, TSV, JSONL, JSON, MD
Size: 449.70 MB
The Pashto Book Summaries and Reviews TTS Corpus, prepared by the Digital Pashto Association and PashtoAI, is a single-speaker Pashto speech dataset designed primarily for text-to-speech and related speech technology research. The corpus contains 1,810 source-verified and structurally screened speech segments totaling 5.478 hours of audio. The recordings consist of summaries and reviews of multiple books narrated in Pashto by Fazl Wali, a 30-year-old male speaker from eastern Afghanistan. The source audio was originally published on his YouTube channel, and written permission was obtained for the use and open release of the audio materials. Each audio segment is paired with a corresponding Pashto transcript. The transcripts preserve the original wording, with only whitespace normalization applied. The single-speaker structure, narrative content, and relatively consistent speaking style make the corpus particularly useful for Pashto text-to-speech development, speech synthesis, acoustic modeling, pronunciation modeling, and speaker-specific speech research. The corpus can also support automatic speech recognition, speech-text alignment, phonetic and pronunciation research, audiobook and spoken-content technologies, language modeling, and other Pashto speech and language applications. The corpus is released under the CC0 1.0 Universal Public Domain Dedication, allowing the data to be used, copied, modified, redistributed, and built upon for research, educational, commercial, and other purposes without seeking additional permission, to the extent permitted by law. By providing more than five hours of transcribed single-speaker Pashto narrative speech covering summaries and reviews of multiple books, the corpus aims to expand openly accessible resources for Pashto TTS, speech synthesis, and low-resource language technology development.
Restrictions/Special Constraints
The dataset owner confirms that signed written permission has been obtained from Fazl Wali for use and open release of the contributed recordings and associated materials, including public redistribution, AI/model-training use, and the selected CC0 1.0 Universal Public Domain Dedication, to the extent of the rights held by the contributor and dataset creators. No additional dataset-specific reuse restrictions are imposed on the rights covered by that dedication. Rights in any third-party material outside their control are not granted by this statement.
Forbidden Usage
No additional dataset-specific prohibitions are imposed on rights covered by the selected CC0 dedication. Applicable law and any third-party rights remain unaffected.
Ethical Review
The dataset owner confirms signed written permission from Fazl Wali for use and open release of the contributed recordings and associated materials, including public redistribution, AI/model-training use, and the selected CC0 release within the contributor’s rights. No independent ethics certification is asserted. Technical and structural checks do not constitute a comprehensive listening, privacy, or third-party-content review; these limitations remain applicable.
Intended Use
Pashto text-to-speech and speech synthesis, automatic speech recognition, speech-text alignment, pronunciation and phonetic research, and speech-language research under the selected CC0 dedication, to the extent of the rights held by the contributors and dataset creators.
The Pashto Book Summaries and Reviews TTS Corpus, prepared by the Digital Pashto Association and PashtoAI, is a Pashto narrative speech dataset containing 1,810 speech segments totaling 5.478 hours. The recordings originate from Pashto book summaries and reviews and are paired with corresponding Pashto transcripts. The corpus is designed to support research in text-to-speech data preparation, speech synthesis, automatic speech recognition, speech-text alignment, pronunciation modeling, phonetic analysis, and other Pashto speech technologies.
The source collection contained 3,377 transcript entries. During preparation, 1,560 entries were excluded based on source inventory criteria and seven additional entries were excluded during packaging, leaving 1,810 retained clips. Original source materials and detailed exclusion records are preserved for provenance and reproducibility.
Long source recordings were fully decoded before segmentation. Audio was cut according to source transcript timestamps rounded to sample boundaries and stored as 16-bit PCM FLAC. Original sample rates and channel configurations were preserved. No resampling, denoising, gain adjustment, or other audio enhancement was applied. Every retained FLAC file was subsequently decoded and compared sample-for-sample as part of technical validation.
The accompanying manifests preserve the original transcript text, normalized text, source filenames, timestamps, hashes, and processing information. Transcript normalization is limited primarily to whitespace normalization, preserving the original linguistic wording.
The source is identified as Fazl Wali / book_readers and belongs to the book-summary domain. Source annotations and available speaker information are retained in the manifests. However, speaker identity, dialect classification, and related metadata have not been independently verified as part of this processing workflow.
The corpus has undergone source verification and structural screening, but these checks should not be interpreted as a new human listening or linguistic certification. No additional listening-based verification, semantic alignment certification, word-boundary validation, or detailed phonetic review was performed. Researchers should therefore conduct additional listening and alignment validation before using the corpus for high-quality production TTS or other applications requiring highly accurate speech-text correspondence.
As stated in the dataset description, written permission was obtained for use and open release of the audio materials. The selected license is the CC0 1.0 Universal Public Domain Dedication, limited to rights held by the contributors and dataset creators; it does not grant rights in the books discussed or other third-party material. The corpus package includes README documentation and metadata describing provenance, processing, exclusions, timestamps, and checksums to support transparent and reproducible use of the resource.
As a relatively consistent narrative-speech collection, the corpus has particular potential for Pashto TTS and speech-synthesis research after appropriate transcript and alignment validation and review of any third-party rights relevant to the intended use. It may also support ASR, audiobook and spoken-content technologies, acoustic modeling, pronunciation research, language modeling, and development of additional Pashto speech resources.
