Datasets | Mozilla Data Collective

Central Kanuri (kau) speech corpus: 3,576 recordings, 12.04 hours of approved audio, 24 contributors, supporting low-resource ASR research.

Central Kanuri Speech Dataset

License: CC-BY-NC-4.0

Locale: knc

Task: ASR

Format: WAV, TSV

Size: 3.35 GB

FHIBE

FHIBE Downsampled

Lower-resolution FHIBE images and pixel-level annotations, resized to fit within a 2048-pixel longest side.

License: FHIBE Term of Use

Locale: en

Task: CV

Format: PNG, JSON, CSV

Size: 44.04 GB

This ASR dataset comprises 10 hours of natural speech in the Javanese-Lumajang dialect in East Java, Indonesia and contains code-switching into Indonesian.

ASR Javanese-Lumajang Dialect

License: CC-BY-NC-SA-4.0

Locale: jav

Task: ASR

Format: OPUS, TSV

Size: 81.38 MB

Rohingya Language

Rohingya Hanifi–Rohingyalish–English Lexicon

15,926 English–Rohingya lexical pairs in Rohingyalish and generated Hanifi script for transliteration, education, and NLP research.

License: CC-BY-4.0

Locale: rhg, en

Task: NLP

Format: JSONL, CSV, MD

Size: 333.79 KB

A collection of read speech recordings in Pashto.

Common Voice Scripted Speech 27.0 - Pashto

License: CC0-1.0

Locale: ps

Task: ASR

Format: MP3

Size: 97.70 GB

This dataset contains various traffic signs and public facilities located along roadways, Magelang, Central Java Province, Indonesia.

IND-Magelang Traffic Signs and Facilities

License: CC-BY-NC-SA-4.0

Locale: ind

Task: CV

Format: JPEG, WEBM, JSON

Size: 118.36 MB

Bahasa Bahana

BAHANA: Parallel Corpus IND-ENG-ARB

The dataset covers a parallel corpus of 5,174 lines of literal translation text in Indonesian, English and Arabic.

License: CC-BY-SA-4.0

Locale: ind,eng,arb

Task: MT

Format: TSV

Size: 437.30 KB

A Northern Pashto ASR dataset with approximately 10 hours of audio with transcriptions.

🚧 Northern Pashto Corpus for Speech Recognition 🚧

License: MDC DLA-1.0

Locale: pbu

Task: ASR

Format: MP3, TSV

Size: 514.50 MB

MirasAICompensated

English 10 Hours Medical Speech Evaluation Dataset (Synthetic)

A 10-hour English medical speech dataset designed for evaluating and improving speech recognition systems in healthcare-related scenarios.

License: MirasAI Data License

Locale: eng

Task: TTS

Format: PARQUET, CSV, JSON, SQL, MD

Size: 922.38 MB

A Tagalog speech recognition dataset consisting of 45+ hours of transcribed conversations recorded via voice memos on WhatsApp and Telegram.

🚧 Tagalog Voice Notes ASR Dataset (45 hours) 🚧

License: MDC DLA-1.0

Locale: tl

Task: ASR

Format: MP3, TSV

Size: 1.24 GB

Thirteen Common Voice Scripted Speech 26.0 locale releases bundled for the LRAC 2.0 Challenge 2026.

LRAC 2.0 Common Voice 26.0 Bundle

License: CC0-1.0

Locale: mul

Task: OTH

Format: MP3, TSV

Size: 209.57 GB

20 hours of studio-quality read Tagalog speech in the health domain (texts from openly-licensed health articles), all from a single female speaker.

🚧 Tagalog Health Domain TTS 20 Hours Single Speaker (Female) 🚧

License: MDC DLA-1.0

Locale: tl

Task: TTS

Format: WAV, TSV

Size: 6.56 GB

25 hours of studio-quality read Tagalog speech in the health domain (texts from openly-licensed health articles), all from a single male speaker.

🚧 Tagalog Health Domain TTS 25 Hours Single Speaker (Male) 🚧

License: MDC DLA-1.0

Locale: tl

Task: TTS

Format: WAV, TSV

Size: 6.74 GB

A dataset of 477 photos covering public spaces across Blitar, Surabaya, Yogyakarta, Surakarta, and Ngawi, Java Island, Indonesia.

Public Space and Sign in Javanese-Speaking Regions

License: CC-BY-NC-SA-4.0

Locale: ind

Task: CV

Format: JPEG, WEBM, JSON

Size: 912.96 MB

TidyVoice2026 Challenge

TidyVoiceM_ASV

TidyVoiceM_ASV (or Tidy-M in the paper) has 212,617 speakers in 81 languages.

License: CC0-1.0

Locale: mul

Task: OTH

Format: WAV

Size: 641.76 GB

A collection of read speech recordings in Fang.

Common Voice Scripted Speech 27.0 - Fang

License: CC0-1.0

Locale: fan

Task: ASR

Format: MP3

Size: 198.88 MB

A collection of read speech recordings in Dawoodi.

Common Voice Scripted Speech 27.0 - Dawoodi

License: CC0-1.0

Locale: dmk

Task: ASR

Format: MP3

Size: 206.93 MB

A collection of read speech recordings in Bashkir.

Common Voice Scripted Speech 27.0 - Bashkir

License: CC0-1.0

Locale: ba

Task: ASR

Format: MP3

Size: 5.10 GB

A collection of read speech recordings in Dutch.

Common Voice Scripted Speech 27.0 - Dutch

License: CC0-1.0

Locale: nl

Task: ASR

Format: MP3

Size: 3.17 GB

A collection of read speech recordings in Slovak.

Common Voice Scripted Speech 27.0 - Slovak

License: CC0-1.0

Locale: sk

Task: ASR

Format: MP3

Size: 1.13 GB

A collection of read speech recordings in Macedonian.

Common Voice Scripted Speech 27.0 - Macedonian

License: CC0-1.0

Locale: mk

Task: ASR

Format: MP3

Size: 1.02 GB

A collection of read speech recordings in Tatar.

Common Voice Scripted Speech 27.0 - Tatar

License: CC0-1.0

Locale: tt

Task: ASR

Format: MP3

Size: 836.51 MB

A collection of read speech recordings in Greek.

Common Voice Scripted Speech 27.0 - Greek

License: CC0-1.0

Locale: el

Task: ASR

Format: MP3

Size: 772.78 MB