datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
VaaniVAANI is an India-representative multi-modal multi-lingual dataset.
The current version (phase 1- 80 districts, phase 2- 85 districts) contains ~31278 hours of spontaenous,image-prompted speech by 156K speakers across 165 districts, talking about 288K images covering 105 languages.
From this audio data, 2,122 hours of transcribed data(text) is available, spanning almost evenly across the 165 districts.
Project Vaani, by IISc, Bangalore and ARTPARK, is capturing the true diversity of India’s… See the full description on the dataset page: https://huggingface.co/datasets/ARTPARK-IISc/Vaani.ASMR-Archive-Processed
ASMR-Archive-Processed (WIP)
Update (2026-04-03): This dataset has reached the Hugging Face Public Storage Limit. After contacting support, we were informed that the only option is to pay for a storage expansion. Consequently, updates to this dataset are now suspended.
Work in Progress — expect breaking changes while the pipeline and data layout stabilize.
This dataset contains ASMR audio data sourced from DeliberatorArchiver/asmr-archive-data-01 and… See the full description on the dataset page: https://huggingface.co/datasets/OmniAICreator/ASMR-Archive-Processed.Earnings22-Cleaned-AA
Earnings22-Cleaned-AA
Quick links: AA Speech-to-Text Leaderboard | AA-WER v2.0 article
Earnings22-Cleaned-AA is a cleaned subset of the English Earnings-22 test data from esb/datasets, a corpus of corporate earnings calls from global companies with speakers of many different nationalities and accents. This cleaned subset is the Earnings-22 portion included in AA-WER v2. We manually reviewed and corrected errors in the original ground-truth transcriptions to ensure fairer evaluation… See the full description on the dataset page: https://huggingface.co/datasets/ArtificialAnalysis/Earnings22-Cleaned-AA.dialectal-arabic-lahgtna-v2
Dialectal Arabic Lahgtna v2
Large-scale multi-dialect Arabic speech dataset — 3,000+ hours across 13 Arabic dialects — for training and evaluating dialectal Arabic ASR systems. Part of the Lahgtna (لهجتنا) project for dialect-aware Arabic speech AI.
Dataset Summary
~611K utterances / 3,000+ hours of transcribed dialectal Arabic speech
**13 Arabic dialects **, labeled per utterance
16 kHz mono audio
Transcripts written in authentic dialectal orthography (not… See the full description on the dataset page: https://huggingface.co/datasets/oddadmix/dialectal-arabic-lahgtna-v2.central-kurdish-pseudolabel
Central Kurdish → English Pseudo-Labeled Speech Translation Corpus
Dataset Summary
This repository contains a large-scale pseudo-labeled speech translation corpus for Central Kurdish (Sorani Kurdish).
The dataset was automatically generated using a pipeline composed of:
Speech segmentation
Automatic Speech Recognition (ASR)
Machine Translation (MT)
The objective is to provide training data for end-to-end Speech-to-Text Translation (S2TT) in a language with very… See the full description on the dataset page: https://huggingface.co/datasets/aranemini/central-kurdish-pseudolabel.voicehub-arena-seed-tts-eval
VoiceHub Arena — full English Seed-TTS-Eval
35,904 synthesized WAV files: 33 model families × the same 1,088 target texts.
The campaign completed on 15 September 2026 on one NVIDIA A100-SXM4 40 GB.
All 198 shards and every WAV SHA256 were verified after backup.
Interactive leaderboard and all audio samples
· Source repository (access required).
Contents
audio_shards/<model>.tar: 33 WebDataset shards, each containing 1,088 original WAVs and matching JSON metadata.… See the full description on the dataset page: https://huggingface.co/datasets/kadirnar/voicehub-arena-seed-tts-eval.MASC-Arabic
MASC Arabic Dataset Card
Dataset Summary
MASC is a dataset that contains 1,000 hours of speech sampled at 16 kHz and crawled from over 700 YouTube channels.
The dataset is multi-regional, multi-genre, and multi-dialect intended to advance the research and development of Arabic speech technology with a special emphasis on Arabic speech recognition.
How to use
The datasets library allows you to load and pre-process your dataset in pure Python, at scale. The… See the full description on the dataset page: https://huggingface.co/datasets/MohamedRashad/MASC-Arabic.dialectal-arabic-voices
Dialectal Arabic Voices
An expanding collection of Arabic audio from YouTube, SoundCloud, and other sources. Currently labelled Palestinian Arabic (ps).
59,812 recordings · approximately 10,634.4 hours · 556.22 GB
Column
Description
audio
Original audio, embedded in the Parquet file
transcript_text
Empty; ASR transcripts are stored in a separate private dataset
language
Dialect code: ps (Palestinian)
source
Original channel or account name
Audio retains its… See the full description on the dataset page: https://huggingface.co/datasets/moaead/dialectal-arabic-voices.arabic-audio-collection-algerian-loubna-stories
Loubna Stories Arabic Speech Dataset
Dataset Summary
The Loubna Stories Arabic Speech Dataset is a large-scale, first-of-its-kind Arabic speech corpus containing approximately 237 hours of speech recordings and corresponding transcripts.
What distinguishes this dataset as a pioneering resource in Arabic language technology is its comprehensive inclusion of rich non-verbal transcriptions. Alongside the spoken Arabic text, the transcripts meticulously capture… See the full description on the dataset page: https://huggingface.co/datasets/oddadmix/arabic-audio-collection-algerian-loubna-stories.Vaani-transcription-partThis dataset is part of the Vaani dataset and consists of only transcribed speech data. It has a total duration of 2041.54 hours, covering 59 languages.
This table represents the audio and transcription duration data for various languages.
Language
Angami
Angika
Ao
Assamese
Awadhi
Bajjika
Bearybashe
Bengali
Bhili
Bhojpuri
Bundeli
Chakhesang
Chakma
Chhattisgarhi
English
Garhwali
Garo
Gondi
Gujarati
Halbi
Haryanvi
Hindi
IduMishmi
Kannada
Kashmiri
Karbi
Khariboli
Khortha
Kokborok
Konkani… See the full description on the dataset page: https://huggingface.co/datasets/ARTPARK-IISc/Vaani-transcription-part.linto-dataset-audio-ar-tn
LinTO DataSet Audio for Arabic Tunisian A collection of Tunisian dialect audio and its annotations for STT task
This is the first packaged version of the datasets used to train the Linto Tunisian dialect with code-switching STT
(linagora/linto-asr-ar-tn).
Dataset Summary
Dataset composition
Sources
Data Table
Data sources
Content Types
Languages and Dialects
Example use (python)
License
Citations
Dataset Summary
The LinTO DataSet Audio for Arabic Tunisian is a diverse… See the full description on the dataset page: https://huggingface.co/datasets/linagora/linto-dataset-audio-ar-tn.mgb2-arabic
MGB-2: Arabic Multi-Dialect Broadcast Media Recognition
Dataset Description
Dataset Summary
The Arabic Multi-Genre Broadcast (MGB-2) dataset is a large-scale speech recognition corpus containing 1,200 hours of Arabic broadcast audio from Aljazeera Arabic TV channel. The dataset spans recordings from March 2005 to December 2015 and covers 19 distinct programme series. It was originally created for the MGB-2 Challenge at SLT-2016, focusing on handling dialect… See the full description on the dataset page: https://huggingface.co/datasets/MohamedRashad/mgb2-arabic.VoxPopuli-Cleaned-AA
VoxPopuli-Cleaned-AA
Quick links: AA Speech to Text Leaderboard | AA-WER v2.0 article
VoxPopuli-Cleaned-AA is a cleaned subset of the English VoxPopuli test data from esb/datasets, a speech dataset derived from European Parliament recordings. This cleaned subset is the VoxPopuli portion included in AA-WER v2. We manually reviewed and corrected errors in the original ground-truth transcriptions to ensure fairer evaluation of Speech to Text (STT) models.
This dataset is part of AA-WER… See the full description on the dataset page: https://huggingface.co/datasets/ArtificialAnalysis/VoxPopuli-Cleaned-AA.ar-quran-hadith14books-MSA
ar-quran-hadith14books-MSA
Arabic speech for both primary sources of Islam — Quran and Hadith — plus cleaned general
Modern Standard Arabic, under one construction pipeline and one text convention.
ASR errors on sacred text are not ordinary errors: a plausible-sounding substitution can alter
the meaning of scripture, and because chatbots, search and summarizers increasingly answer from
transcriptions rather than from audio, such an error propagates silently. Quranic recitation… See the full description on the dataset page: https://huggingface.co/datasets/Dr-AliGomaa/ar-quran-hadith14books-MSA.NEXUS-temporal_hierarchical_multi-modal
NEXUS: Neural Evolution for eXtensible Universal Semantics Dataset
(Temporal Multimodal Slices)
This dataset is a multi-modal, hierarchical, temporal representation derived from HuggingFaceFV/finevideo. It is designed for streaming training where the primary unit is a 10 ms "slice" that aggregates upward into moments (100 ms), seconds (1 s), experiences (10 s), and minutes (60 s).
It is meant to represent an extensible stream of "experience" as there are… See the full description on the dataset page: https://huggingface.co/datasets/Ardea/NEXUS-temporal_hierarchical_multi-modal.northern-kurdish-raw-audio
Northern Kurdish Raw Audio Collection
Overview
This repository contains a large collection of raw Northern Kurdish (Kurmanji Kurdish) speech recordings gathered from publicly available Kurdish media sources.
The collection was assembled to support research and development in:
Automatic Speech Recognition (ASR)
Speech Translation (ST)
Text-to-Speech (TTS)
Self-supervised Learning (SSL)
Spoken Language Understanding (SLU)
The dataset contains more than 2,000 hours… See the full description on the dataset page: https://huggingface.co/datasets/aranemini/northern-kurdish-raw-audio.ArVoice
ArVoice: A Multi-Speaker Dataset for Arabic Speech Synthesis
Hawau Olamide Toyin, Rufael Marew, Humaid Alblooshi, Samar M. Magdy, Hanan Aldarmaki
{hawau.toyin, hanan.aldarmaki}@mbzuai.ac.ae
ArVoice is a multi-speaker Modern Standard Arabic (MSA) speech corpus with fully diacritized transcriptions, intended for multi-speaker speech synthesis, and can be useful for other tasks such as speech-based diacritic restoration, voice conversion, and deepfake detection.
ArVoice… See the full description on the dataset page: https://huggingface.co/datasets/MBZUAI/ArVoice.ASMR-Archive-Processed-SFW
ASMR-Archive-Processed-SFW
Overview
This dataset is an “educational” subset of the original OmniAICreator/ASMR-Archive-Processed dataset.
We filtered the original dataset to include only records where the nsfw metadata flag is false.
To maintain the randomness and anonymity of the entries, multiple directories were combined and shuffled.
The nsfw tag in the original dataset is inherited from the tags of the original audio works before they were passed through the… See the full description on the dataset page: https://huggingface.co/datasets/noxwano/ASMR-Archive-Processed-SFW.Earnings22-Cleaned-AA-chunked
Earnings22-Cleaned-AA-chunked
Quick links: AA Streaming Speech to Text Leaderboard | Speech to Text methodology
Earnings22-Cleaned-AA-chunked is a chunked version of Earnings22-Cleaned-AA, the cleaned Earnings-22 subset used by Artificial Analysis for streaming Speech to Text evaluation.
The original Earnings-22 data comes from esb/datasets, a corpus of corporate earnings calls. Artificial Analysis manually reviewed and corrected the reference transcripts in the cleaned subset… See the full description on the dataset page: https://huggingface.co/datasets/ArtificialAnalysis/Earnings22-Cleaned-AA-chunked.linto-dataset-audio-ar-tn-augmented
LinTO DataSet Audio for Arabic Tunisian Augmented A collection of Tunisian dialect audio and its annotations for STT task
This is the augmented datasets used to train the Linto Tunisian dialect with code-switching STT linagora/linto-asr-ar-tn.
Dataset Summary
Dataset composition
Sources
Content Types
Languages and Dialects
Example use (python)
License
Citations
Dataset Summary
The LinTO DataSet Audio for Arabic Tunisian Augmented is a dataset that builds on LinTO… See the full description on the dataset page: https://huggingface.co/datasets/linagora/linto-dataset-audio-ar-tn-augmented.librispeech-arpabet-processed
LibriSpeech ARPAbet Processed Dataset
Pre-processed dataset for training ARPAbet phoneme recognition models using CTC loss.
Dataset Description
This dataset is derived from LibriSpeech (train-clean-100 split) with the following preprocessing:
Audio: Resampled to 16kHz, normalized using Wav2Vec2 feature extractor
Labels: Text transcriptions converted to ARPAbet phoneme sequences using CMU Pronouncing Dictionary
Filtering: Samples with out-of-vocabulary words (not in CMU… See the full description on the dataset page: https://huggingface.co/datasets/davidggphy/librispeech-arpabet-processed.common-voice-18-arabic
Dataset Card for Common Voice 18 – Arabic Edition
Dataset Summary
This dataset is an unofficial Arabic-only extraction of Mozilla Common Voice Corpus 18.0, prepared for Automatic Speech Recognition (ASR) research and development.
It is derived from the original Common Voice 18 release and filtered to include Arabic (ar) speech data only, while preserving the original dataset structure, splits, and metadata fields. The dataset consists of validated, unvalidated, and… See the full description on the dataset page: https://huggingface.co/datasets/MohamedRashad/common-voice-18-arabic.arknights_voices_zh
ZH Voice-Text Dataset for Arknights Waifus
This is the ZH voice-text dataset for arknights playable characters. Very useful for fine-tuning or evaluating ASR/ASV models.
Only the voices with strictly one voice actor is maintained here to reduce the noise of this dataset.
12431 records, 25.9 hours in total. Average duration is approximately 7.49s.
id
char_id
voice_actor_name
voice_title
voice_text
time
sample_rate
file_size
filename
mimetype
file_url
char_106_franka_CN_001… See the full description on the dataset page: https://huggingface.co/datasets/deepghs/arknights_voices_zh.simchoir-parquet
FastMSS synthetic multi-speaker meetings - parquet edition
Streaming-friendly parquet shards of the FastMSS synthetic multi-speaker conversational corpus. Each row is one mixture with the audio bytes embedded inline (16 kHz mono WAV) plus per-segment diarization timestamps, per-word transcript and the full lhotse cut as a JSON blob. See fastmss/hf_dataset.py for the schema docstring.
Subsets and splits
debug — splits: train — 1 mixtures, 1.6 min total, 6 unique speakers… See the full description on the dataset page: https://huggingface.co/datasets/arda-argmax/simchoir-parquet.commonvoice-12.0-arabic-voice-converted
Dataset Card for Voice Converted Arabic Common Voice 12.0
This dataset is derived from the Common Voice Arabic Corpus 12.0 and includes automatically diacritized transcriptions and phoneme representations for the original augmented audio data. The recordings feature Arabic text read aloud by users, where the text was initially undiacritized, allowing for potential reading errors. The diacritization and phonemes were generated automatically, resulting in a dataset that is valuable… See the full description on the dataset page: https://huggingface.co/datasets/xmodar/commonvoice-12.0-arabic-voice-converted.arabic-audio-collection-algerian-kahwa-postcast
Kahwa Postcast Arabic Speech Dataset
Dataset Summary
The Kahwa Postcast Arabic Speech Dataset is a large-scale, first-of-its-kind Arabic speech corpus containing approximately 110 hours of speech recordings and corresponding transcripts.
What distinguishes this dataset as a pioneering resource in Arabic language technology is its comprehensive inclusion of rich non-verbal transcriptions. Alongside the spoken Arabic text, the transcripts meticulously capture… See the full description on the dataset page: https://huggingface.co/datasets/oddadmix/arabic-audio-collection-algerian-kahwa-postcast.common_voice_20_armenian
Common Voice 20 - Armenian
This dataset is the Armenian portion of Mozilla's Common Voice 20.0 release,
a massively multilingual collection of transcribed speech intended for speech technology research and development.
Dataset Details
Language: Armenian (hy)
Source: Mozilla Common Voice
Version: 20.0
License: CC0-1.0
nhk-archive-audio-30s
NHK Archives Audio 30s
This is a Japanese speech corpus derived from NHK Archives Audio. Audio from public NHK Archives records was segmented into clips of up to 30 seconds using voice activity detection.
The dataset contains 344,722 accepted clips, totaling 1,661.01 hours. Audio is embedded as 16 kHz mono FLAC. raw_text was transcribed with Whisper large-v3-turbo, and text contains LLM-assisted corrections based on the transcript and available source title and description.
This… See the full description on the dataset page: https://huggingface.co/datasets/KeisukeMiyamoto/nhk-archive-audio-30s.french_tv_media_dataset_2026
Dataset Card : A Multi-Domain Pseudo-Labeled ASR Corpus
Résumé (Abstract)
Ce corpus présente un jeu de données de reconnaissance automatique de la parole (ASR) en langue française, totalisant 97 heures d'audio annoté. Il est dérivé de flux de diffusion (broadcast) issus de France Télévisions, couvrant une diversité de domaines acoustiques et linguistiques (Information, Société, Divertissement, Documentaire, Sport). L'annotation a été réalisée via une méthodologie… See the full description on the dataset page: https://huggingface.co/datasets/Archime/french_tv_media_dataset_2026.CoVoST2-EN-AR
Dataset Description
CoVoST 2 is a large-scale multilingual speech translation corpus based on Common Voice, developed by FAIR. This is the English-to-Arabic portion of the dataset. The original dataset can be found here.
Data Splits (EN-AR)
lang
train
validation
test
EN-AR
289430
15531
15531
AR-EN
2283
1758
1695
Citation
@misc{wang2020covost,
title={CoVoST 2: A Massively Multilingual Speech-to-Text Translation Corpus},
author={Changhan… See the full description on the dataset page: https://huggingface.co/datasets/ymoslem/CoVoST2-EN-AR.