Datasets | Mozilla Data Collective

The dataset includes audio files in English and their translations in Arabic.

CoVoST 2 English - Arabic

License: CC-BY-NC-4.0

Locale: en, ar

Task: MT

Format: MP3, TSV

Size: 12.56 GB

25 hours of Conversational Speech in the Gujarati Language in the Medical Domain

Conversational Speech in Gujarati in the Medical Domain

License: MDC-DLA 1.0

Locale: gu

Task: ASR

Format: WAV, JSON

Size: 6.18 GB

Text to speech dataset for Spanish, male speaker, approximately 1.5 hours of read speech.

Dave 1.0

License: CC0-1.0

Locale: es-ES

Task: TTS

Format: WEBM

Size: 85.24 MB

Tamazight NLP

Tamazight Open Speech Dataset

This dataset provides a parsed, formatted, and ready-to-use Amazigh Voice Dataset. It contains voice recordings and corresponding text transcripts in Standard Moroccan Amazigh (ⵜⴰⵎⴰⵣⵉⵖⵜ ⵜⴰⵏⴰⵡⴰⵢⵜ ⵜⴰⵎⵓⵔⴰⴽⵓⵛⵜ) intended for training Automatic Speech Recognition (ASR) and Text-to-Speech (TTS) models.

License: Apache-2.0

Locale: zgh

Task: ASR

Format: WAV, JSONL

Size: 459.09 MB

Text to speech dataset for Russian, male speaker, approximately 2 hours of read speech.

Denis 1.0

License: CC0-1.0

Locale: ru-RU

Task: TTS

Format: WEBM

Size: 104.52 MB

Kaleem Art Press

Jhoke Publisher Multan’s Saraiki Newspaper Corpus

Jhoke Publisher Multan’s Saraiki Newspaper Corpus (~1.25M tokens) is a normalized UTF-8 collection of Saraiki newspaper.

License: CC-BY-NC-4.0

Locale: skr

Task: NLP

Format: TXT

Size: 2.30 MB

RFE/RL

RFE/RL Pashto (Pakistani) News Text Corpus

Longitudinal Pakistani Pashto news corpus from Radio Mashaal (2010-2026) with over 52,000 articles and 14.5M tokens.

License: CC-BY-NC-SA-4.0

Locale: ps

Task: NLP

Format: TXT

Size: 39.26 MB

MDC Community Concierge

VoxForge - Hebrew

32 minutes (307 utterances) of read speech of ‭עברית‬ (Hebrew).

License: GPL-3.0

Locale: he

Task: ASR

Format: WAV, TSV

Size: 126.98 MB

Text to speech dataset for Russian, male speaker, approximately 2 hours of read speech.

Dmitri 1.0

License: CC0-1.0

Locale: ru-RU

Task: TTS

Format: WEBM

Size: 96.63 MB

25 hours of Conversational Speech in the Gujarati Language in the Finance Domain

Conversational Speech in Gujarati in the Finance Domain

License: MDC DLA-1.0

Locale: gu

Task: ASR

Format: WAV, JSON

Size: 6.20 GB

Text to speech dataset for English, male speaker, approximately 1 hour of read speech.

Joe 1.0

License: CC0-1.0

Locale: en-US

Task: TTS

Format: WEBM

Size: 75.78 MB

RFE/RL

RFE/RL Tajik News Text Corpus

Longitudinal Tajik & Russian news corpus from Radio Ozodi (2000-2026) with over 166,000 articles and 20M tokens.

License: CC-BY-NC-SA-4.0

Locale: tg,ru

Task: NLP

Format: TXT

Size: 145.27 MB

Arabic Common Voice scripted speech 26.0 segment of Male speakers.

Common Voice Scripted Speech 26.0 - Arabic (Male)

License: CC0-1.0

Locale: ar

Task: ASR

Format: MP3, TSV

Size: 235.91 MB

A Hazargi-English parallel corpus derived from transcribed and translated video recordings

🚧 Hazargi-English Parallel Text Corpus 🚧

License: MDC DLA-1.0

Locale: haz, en

Task: MT

Format: TSV

Size: 434.04 KB

2 hours of sentence-aligned speech/text from "Krysař" by Viktor Dyk

LibriVox Czech TTS Female Voice

License: CC0-1.0

Locale: cs

Task: TTS

Format: MP3, TXT, TSV

Size: 178.58 MB

A Khowar multimodal dataset with 10 hours of synchronized video, audio, transcriptions, speech segmentation, and ELAN annotations.

🚧 Khowar Multimodal Dataset 🚧

License: MDC DLA 1.0

Locale: khw, en

Task: NLP

Format: MP4, WAV, EAF

Size: 167.23 GB

Tamir News Agency

Tamir Sindhi News Corpus

The corpus contains 1.1 million tokens from the Tamir Sindhi Newspaper published from the year 2022-2025. The text consists of the complete newspaper content including headlines, editorials, finance news and advertisements. The newspaper published in Karachi, Pakistan on daily basis.

License: CC-BY-NC-SA-4.0

Locale: snd

Task: NLP

Format: TXT

Size: 2.56 MB

RFE/RL

RFE/RL Tatar-Bashkir News Text Corpus

This dataset is longitudinal news corpus for the Tatar and Bashkir languages, sourced from Azatliq Radiosi, from 2001 to 2025, it contains over 105,000 articles (30M tokens).

License: CC-BY-NC-SA-4.0

Locale: tt,ba,ru

Task: NLP

Format: TXT

Size: 102.44 MB

Habib University

Burushaski-English Speech Translation Corpus

A linguistically informed Burushaski speech - English text corpus containing 14,970 short audio recordings paired with corresponding English text translations.

License: CC-BY-NC-SA-4.0

Locale: bsk

Task: MT

Format: WAV, TXT

Size: 921.47 MB

A Hazargi ASR dataset with approximately 10 hours of audio with transcriptions.

🚧 Hazargi Corpus for Speech Recognition 🚧

License: MDC DLA-1.0

Locale: haz

Task: ASR

Format: MP3, TSV

Size: 511.56 MB

Habib University

Burushaski-English Speech Translation Corpus

A linguistically informed Burushaski speech - English text corpus containing 14,970 short audio recordings paired with corresponding English text translations.

License: CC-BY-NC-SA-4.0

Locale: bsk

Task: MT

Format: WAV, TXT

Size: 921.47 MB

A closed-form South-Asian-English-accent subset of Mozilla Common Voice Scripted Speech 26.0 (English, validated). MP3 audio + TSV metadata.

Common Voice Scripted Speech 26.0 - South Asian English (India, Pakistan, Sri Lanka)

License: CC0-1.0

Locale: en

Task: ASR

Format: TSV, MP3

Size: 3.83 GB