Datasets | Mozilla Data Collective

CDS

MALOBA Lingala Speech Corpus (ASR), v3.2

18,767 Lingala audio segments (58.1 h) with original and harmonised transcriptions, speaker-disjoint splits, clean test set for Whisper evaluation.

License: NOODL-1.0

Locale: ln

Task: ASR

Format: Parquet

Size: 4.45 GB

CLEAR Global

Hausa Multispeaker Read Speech (6.7h)

4,986 Hausa read-speech audio clips (~6.7 hours) with English source sentences, from CLEAR Global's Gamayun mini-kit.

License: CC-BY-NC-4.0

Locale: hau

Task: ASR

Format: MP3, TSV

Size: 214.61 MB

CDS

MALOBA French-Lingala translation pairs (h2)

4,874 French → Lingala sentence pairs (399 human translations + question pairs), harmonised Lingala spelling.

License: NOODL-1.0

Locale: ln

Task: MT

Format: JSONL

Size: 338.41 KB

Community

IDN - Spanish Corpus Learner

A corpus of read speech by learners of Spanish from Indonesia.

License: CC-BY-SA-4.0

Locale: es

Task: CALL

Format: WAV, WEBM, TSV

Size: 1.26 GB

Key Laboratory of Ethnic Language Intelligent Analysis and Security Governance of MOE

UyghurDF

Preview subset of UyghurDF, a multi-generator Uyghur speech deepfake dataset for detection and cross-generator generalization research.

License: CC-BY-4.0

Locale: ug

Task: OTH

Format: WAV

Size: 6.94 MB

Pashto AI

Pashto Audio-Visual Speech Corpus

Pashto audio-visual corpus with 9,421 recordings (12.67 hours), pairing speech, mouth-region video and text for lip reading, ASR and multimodal AI.

License: CC0-1.0

Locale: ps

Task: ASR

Format: WEBM, TSV, JSON, JSONL, MD, TXT

Size: 1.33 GB

Community

Yaayuwee Gbaya Medicinal Plants 18

Dataset de 18 plantes médicinales Gbaya (Bertoua) avec 11 photos originales et noms en Gbaya, français et scientifique.

License: CC-BY-4.0

Locale: gya

Task: CV

Format: JPG, CSV, MD

Size: 35.85 MB

Mozilla Foundation

Faculty Atlas: US Faculty Job Postings (2026-10-01)

Point-in-time metadata for 23,265 US faculty job postings from 1,636 higher-education employers, captured 2026-10-01.

License: CC-BY-4.0

Locale: en

Task: N/A

Format: CSV, JSON, MD

Size: 4.17 MB

Community

Synthetic Spanish–English Code-Switching Speech (DS-3)

2,233 synthetic Spanish–English code-switched utterances (5.78 h), synthetic TTS voices only, made for Lost in Transcription.

License: GPL-3.0

Locale: es,en

Task: ASR

Format: WAV, JSONL

Size: 550.01 MB

Community

Large Javanese ASR training data set (OpenSLR 35), 8,000-utterance subset

8,000 read Javanese utterances (12.7 h) with transcripts: a subset of OpenSLR 35, collected by Google, Reykjavik University and Universitas Gadjah Mada.

License: CC-BY-SA-4.0

Locale: jv

Task: ASR

Format: FLAC, TSV

Size: 713.26 MB

Community

High quality TTS data for Javanese (OpenSLR 41)

5,822 high-quality read Javanese utterances (7.0 h) with transcripts from OpenSLR 41, collected by Google with Universitas Gadjah Mada.

License: CC-BY-SA-4.0

Locale: jv

Task: ASR

Format: WAV, TSV

Size: 1.76 GB

Community

Kanuri Lexicon

A structured Kanuri lexical resource containing headwords, grammatical labels, English definitions, and source entry information.

License: CC-BY-NC-4.0

Locale: kau

Task: N/A

Format: CSV, JSONL, TXT, MD

Size: 577.46 KB

Community

Ukrainian–Russian–Surzhyk Read Speech

3,117 read-speech recordings containing Ukrainian, Russian, and Surzhyk, with transcripts and anonymous speaker IDs

License: SURZHYK-TERMS-0.1.0

Locale: uk, ru

Task: ASR

Format: WAV, TSV, MD

Size: 240.92 MB

Community

Synthetic Javanese–Indonesian Conversational Speech

13,469 synthetic Javanese–Indonesian conversational utterances (~61 h) made with Gemma 4 text and VoxCPM2 speech for Lost in Transcription.

License: CC-BY-SA-4.0

Locale: jv,id

Task: ASR

Format: WAV, TSV

Size: 5.25 GB

CDS

MALOBA Lingala OCR transcriptions, harmonised text (v1.1)

1,282 harmonised Lingala OCR text lines (104 pages, 535 NLLB units), text only, OCR test set excluded.

License: NOODL-1.0

Locale: ln

Task: NLP

Format: JSONL

Size: 391.72 KB

CLEAR Global

TICO-19 (Covid-domain) Speech – Congolese Swahili

TICO-19 COVID-19/medical domain sentences read aloud in Congolese Swahili, recorded by CLEAR Global in 2021

License: CC-BY-NC-4.0

Locale: swc

Task: ASR

Format: MP3, TSV

Size: 190.76 MB

CLEAR Global

Read Speech in Congolese Swahili (11h)

5,000 Congolese Swahili read-speech audio clips (~11 hours) with French translations, from CLEAR Global's Gamayun mini-kit.

License: CC-BY-NC-SA-4.0

Locale: swc

Task: ASR

Format: MP3, TSV

Size: 144.25 MB

FHIBE

FHIBE Metadata

Metadata files for the FHIBE dataset variants.

License: FHIBE Term of Use

Locale: en

Task: CV

Format: PNG, JSON, CSV, TXT

Size: 3.88 GB

FHIBE

FHIBE Full Resolution

Original FHIBE images and pixel-level annotations at full resolution.

License: FHIBE Term of Use

Locale: en

Task: CV

Format: PNG, JSON, CSV

Size: 135.58 GB

FHIBE

FHIBE Face Crop Only

FHIBE face crops created directly from full-resolution images.

License: FHIBE Term of Use

Locale: en

Task: CV

Format: PNG, JSON, CSV

Size: 10.63 GB

FHIBE

FHIBE Face Crop Align

Cropped and aligned FHIBE face images with key landmarks normalized across samples.

License: FHIBE Term of Use

Locale: en

Task: CV

Format: PNG, JSON, CSV

Size: 2.79 GB

FHIBE

FHIBE Downsampled Overlays

FHIBE Downsampled images with bounding boxes, masks, or keypoints overlaid for visualization.

License: FHIBE Term of Use

Locale: en

Task: CV

Format: PNG

Size: 128.59 GB

Community

Italian female voice reading Italian Penal Code

Italian Female Voice, 60yo, reading the italian penal code

License: Apache-2.0

Locale: ITA

Task: TTS

Format: .wav

Size: 3.71 GB

Brett Williams Studios LLCCompensated

🚧 Hitch Hiker 'Mirror Suit' Robotics Data 🚧

High-specularity optical edge-case dataset (425 assets) for benchmarking computer vision and autonomous navigation against severe specular reflections.

License: LICENSE.md

Locale: en

Task: CV

Format: NEF, JPEG

Size: 4.47 GB

All Datasets

Compensated datasets

English 10 Hours Medical Speech Evaluation Dataset (Synthetic)

English Medical Speech Dataset (Synthetic)

Single Speaker Urdu TTS dataset (Female Speaker, South Punjab)

MALOBA Lingala Speech Corpus (ASR), v3.2

Hausa Multispeaker Read Speech (6.7h)

MALOBA French-Lingala translation pairs (h2)

IDN - Spanish Corpus Learner

UyghurDF

Pashto Audio-Visual Speech Corpus

Yaayuwee Gbaya Medicinal Plants 18

Faculty Atlas: US Faculty Job Postings (2026-10-01)

Synthetic Spanish–English Code-Switching Speech (DS-3)

Large Javanese ASR training data set (OpenSLR 35), 8,000-utterance subset

High quality TTS data for Javanese (OpenSLR 41)

Kanuri Lexicon

Ukrainian–Russian–Surzhyk Read Speech

Synthetic Javanese–Indonesian Conversational Speech

MALOBA Lingala OCR transcriptions, harmonised text (v1.1)

TICO-19 (Covid-domain) Speech – Congolese Swahili

Read Speech in Congolese Swahili (11h)

FHIBE Metadata

FHIBE Full Resolution

FHIBE Face Crop Only

FHIBE Face Crop Align

FHIBE Downsampled Overlays

Italian female voice reading Italian Penal Code

🚧 Hitch Hiker 'Mirror Suit' Robotics Data 🚧