datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
librispeech_asr
Dataset Card for librispeech_asr
Dataset Summary
LibriSpeech is a corpus of approximately 1000 hours of 16kHz read English speech, prepared by Vassil Panayotov with the assistance of Daniel Povey. The data is derived from read audiobooks from the LibriVox project, and has been carefully segmented and aligned.
Supported Tasks and Leaderboards
automatic-speech-recognition, audio-speaker-identification: The dataset can be used to train a model for Automatic… See the full description on the dataset page: https://huggingface.co/datasets/openslr/librispeech_asr.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.open-yap-1k
Open Yap 1K: 1,000 hours of full-duplex natural conversation, free for commercial use
Today we're releasing Open Yap 1K: 1,000 hours of dual-channel English conversation, capturing how people speak together naturally in real-world environments recorded in 48kHz. The dataset ships free for both commercial and research use.
The sample on the Hugging Face Hub - 8.9 hours, 16 conversations, CC-BY-4.0, listenable in the dataset viewer.
The full corpus - 1,000 hours, 1,602… See the full description on the dataset page: https://huggingface.co/datasets/TheAgenticDataCompany/open-yap-1k.omnilingual-asr-corpus
Meta Omnilingual ASR Corpus
The Omnilingual ASR Corpus is a collection of spontaneous speech recordings and their transcriptions for 348 under-served languages. The corpus was collected as part of Meta FAIR’s Omnilingual ASR project (blog, model, paper) for the purposes of training automatic speech recognition (ASR) and spoken language identification models.
Data schema
{
`language`: "lij_Latn",
`iso_639_3`: "lij",
`iso_15924`: "Latn",
`glottocode`:… See the full description on the dataset page: https://huggingface.co/datasets/facebook/omnilingual-asr-corpus.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.DeepDialogue-orpheus
DeepDialogue-orpheus
DeepDialogue-orpheus is a large-scale multimodal dataset containing 40,150 high-quality multi-turn dialogues spanning 41 domains and incorporating 20 distinct emotions with coherent emotional progressions. This repository contains the Orpheus variant of the dataset, where speech is generated using Orpheus, a state-of-the-art TTS model that infers emotional expressions implicitly from text.
🚨 Important Notice
This dataset is large (~180GB) due to… See the full description on the dataset page: https://huggingface.co/datasets/SALT-Research/DeepDialogue-orpheus.open-bible-speech-african
Open Bible Resources — African Languages
Spoken-audio Bible recordings aligned to verse-level text for 19 African languages —
roughly 1,741 hours of audio across ~552,907 audio–text pairs (~357 GB).
This dataset is the African-language subset of
davidguzmanr/open-bible-resources,
re-hosted here by AfriSpeech to make the African
languages easy to find and use on their own. The audio and text are unchanged from the
source; only the non-African configurations have been removed. All… See the full description on the dataset page: https://huggingface.co/datasets/AfriSpeech/open-bible-speech-african.open-bible
OpenBibleTTS
OpenBibleTTS is a large-scale, multilingual speech corpus for low-resource text-to-speech (TTS), spanning 37 underrepresented languages across five regions. It contains ~3,469 hours of aligned, verse-level read speech and 1,121,956 utterances, derived from the Open Bible platform and released under a permissive license.
Alignment pipeline: https://github.com/davidguzmanr/open-bible-resources
Source: Open Bible (CC BY-SA)
Languages
Africa (19), South… See the full description on the dataset page: https://huggingface.co/datasets/multilingual-tts/open-bible.yt-danish-public-v2ASR-datasets-ptbr
📚 Datasets de Áudio em Português (PT-BR)
Este repositório reúne diversos corpora públicos de fala em português do Brasil, combinados em um único dataset para facilitar treinamentos e pesquisas em ASR (Automatic Speech Recognition).
O objetivo é fornecer um recurso amplo, padronizado e de fácil acesso para a comunidade.
📂 Datasets Integrados
A tabela abaixo lista todos os datasets incluídos, com suas informações:
Dataset
Config Name
TOTAL
train
test
validation… See the full description on the dataset page: https://huggingface.co/datasets/opedromartins/ASR-datasets-ptbr.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.lahgtna-v3-small
Lahgtna — Dialect-Balanced Arabic ASR (v3 small)
A dialect-balanced multi-dialect Arabic speech-recognition corpus: 54,600 clips /
267.3 hours across 13 Arabic dialects, 16 kHz mono. Each dialect is evenly
represented — 4,000 train + 200 test clips per dialect — so models and
evaluations aren't skewed toward high-resource dialects (e.g. Egyptian/Gulf).
Used to train the oddadmix v2 dialectal-ASR model family.
Duration by dialect
Dialect
Train (h)
Train clips… See the full description on the dataset page: https://huggingface.co/datasets/oddadmix/lahgtna-v3-small.open-asr-leaderboard-multilingual-datasets
ASR Leaderboard Datasets
This repository contains test splits from multiple speech corpora, including FLEURS, Common Voice (MCV), and Multilingual LibriSpeech (MLS).
How to Load
To load a specific subset, use load_dataset with the corresponding config_name in the format <set>_<lang>.
from datasets import load_dataset
# Load the FLEURS dataset for Bulgarian
fleurs_bg = load_dataset("nithinraok/asr-leaderboard-datasets", "fleurs_bg")
print(fleurs_bg)
# Load the… See the full description on the dataset page: https://huggingface.co/datasets/hf-audio/open-asr-leaderboard-multilingual-datasets.OmnilingualASR-retrieval
Omnilingual ASR speech-text retrieval (MTEB)
Read speech paired with its human transcription, for languages that no existing
MTEB audio task covers.
Source: facebook/omnilingual-asr-corpus at revision 8648ba8, cc-by-4.0, official
test split. Recordings are re-encoded from FLAC to Opus at 16 kHz. Repeated
transcripts are dropped, since one would otherwise be relevant to several
recordings while only one is marked correct.
Built by… See the full description on the dataset page: https://huggingface.co/datasets/vnahata/OmnilingualASR-retrieval.CrawlSinger-OS
CrawlSinger-OS
CrawlSinger-OS is a large-scale, open-source singing corpus constructed for
score-native singing voice synthesis. It contains more than 2,300 hours of
processed singing data from multiple public song and singing collections, with
a unified annotation scheme for lyrics, MIDI pitches, symbolic note values,
lyric-to-note alignment, and global tempo.
VocalRender paper
VocalRender code
VocalRender checkpoints
Why CrawlSinger-OS
Modern singing… See the full description on the dataset page: https://huggingface.co/datasets/pymaster/CrawlSinger-OS.UltiMed-ASR-FR-v1
UltiMed-ASR-FR-v1
A large, fully documented French medical speech dataset for evaluating or training models, plus an open recipe to rebuild it in any language or topic.
Contents
Changelog
What is UltiMed-v1
Why I made UltiMed
Who made UltiMed
Quick start (TL;DR)
How I made UltiMed
Sources
Breakdown by source
Text generation (LLM)
Audio synthesis (TTS)
Hardware and conditions
Repository layout on the Hub
Row format and the two text fields
Loading and training… See the full description on the dataset page: https://huggingface.co/datasets/Olicorne/UltiMed-ASR-FR-v1.ace-opencpop-segments
Citation Information
@misc{shi2024singingvoicedatascalingup,
title={Singing Voice Data Scaling-up: An Introduction to ACE-Opencpop and ACE-KiSing},
author={Jiatong Shi and Yueqian Lin and Xinyi Bai and Keyi Zhang and Yuning Wu and Yuxun Tang and Yifeng Yu and Qin Jin and Shinji Watanabe},
year={2024},
eprint={2401.17619},
archivePrefix={arXiv},
primaryClass={cs.SD},
url={https://arxiv.org/abs/2401.17619},
}
MonsoonASR-Open-ASR-leaderboard-en-IN
Voice Arena Monsoon en-IN (public test)
Part of the Open ASR Leaderboard, in the main board's default column set, so it contributes to the headline Average WER for every model listed.
A conversational Indian English ASR test set that records who was speaking, not only what
was said. Every clip carries twelve speaker attributes — gender, age, native district
and state, education, occupation, income band, handset — so a difference between two
systems can be traced to a group of… See the full description on the dataset page: https://huggingface.co/datasets/VoiceArena/MonsoonASR-Open-ASR-leaderboard-en-IN.HearInContextEnglish | 中文
HearInContext
A Benchmark for Implicit Context in Speech Recognition
Illustrative example: the same spoken request is disambiguated as flour or flower by different assistant histories. The dialogue and waveform are illustrative.
Same audio. Different contexts. Different meanings.
HearInContext is a Mandarin–English contextual speech recognition benchmark. It pairs the same audio with dialogue histories supporting different meanings to evaluate… See the full description on the dataset page: https://huggingface.co/datasets/OPPOer/HearInContext.omnilingual-asr-corpus
Meta Omnilingual ASR Corpus
The Omnilingual ASR Corpus is a collection of spontaneous speech recordings and their transcriptions for 348 under-served languages. The corpus was collected as part of Meta FAIR’s Omnilingual ASR project (blog, model, paper) for the purposes of training automatic speech recognition (ASR) and spoken language identification models.
Data schema
{
`language`: "lij_Latn",
`iso_639_3`: "lij",
`iso_15924`: "Latn"… See the full description on the dataset page: https://huggingface.co/datasets/KathleenKunLiu/omnilingual-asr-corpus.parliament
Parliament
Parliament is an OpenFormosa Traditional Chinese speech dataset derived from the
zh_tw split of disco-eth/WorldSpeech.
It contains audio clips and human transcripts from Taiwan Legislative Yuan IVOD
parliamentary proceedings.
This release keeps only rows that passed the Taiwan-OmniData / FineWeb2-style
text filtering pipeline. Audio is preserved from the upstream dataset and cast
as a Hugging Face Audio(sampling_rate=24000) feature.
Dataset Summary… See the full description on the dataset page: https://huggingface.co/datasets/OpenFormosa/parliament.open-large-bengali-asr-data
Open Large Bengali ASR Data
This is a collection of publicly available ASR data for Bengali. It contains 5000 hours of audio. We have a filtering column called is_better to filter good-quality audio from the corpus. It is set based on the wer between original transcription and prediction taken from a Bengali-Wav2Vec2 model and word-per-second (wps).
Datasets:
commonvoice
ohun
ohùn — Igbo · Yorùbá · Hausa · Pidgin speech corpus
ohùn (Yorùbá for voice) merges the three WaZoBiaSpeech corpora published by
Africanvoice into a single repository, so all
three of Nigeria's major languages can be pulled from one place.
Audio is byte-identical to the sources — this repo re-registers the very same
objects, it does not re-encode anything.
Contents
718,336 utterances · 1,035 GB of audio across four languages.
config
split
rows
shards
size… See the full description on the dataset page: https://huggingface.co/datasets/kapturecx/ohun.or_in_datasetopen-large-bengali-asr-data
Open Large Bengali ASR Data
This is a collection of publicly available ASR data for Bengali. It contains 5000 hours of audio. We have a filtering column called is_better to filter good-quality audio from the corpus. It is set based on the wer between original transcription and prediction taken from a Bengali-Wav2Vec2 model and word-per-second (wps).
Datasets:
commonvoice
openslr
madasr
shrutilipi
flerus
kathbath
indictts
ucla
gali
librispeech-pc-44khz-opus
LibriSpeech-PC 44kHz Opus
Summary
This dataset is a high-quality audio replacement variant of Librispeech PC. It preserves the row identity and text fields while replacing audio content from the source audio with the highest available quality (usually mp3 128kpbs) which is then encoded as Opus (64 kbps). Sampling rate is increased from 16khz up to 48khz depending the on source audio.
LibriSpeech-PC is a merge of openslr/librispeech_asr audio metadata with SLR145… See the full description on the dataset page: https://huggingface.co/datasets/mythicinfinity/librispeech-pc-44khz-opus.youtube_transcriptions
Dataset Description
A speech dataset of Uzbek language audio clips sourced from YouTube videos. Audio segments were extracted, separated by speaker using vocal isolation, and transcribed using Google's Gemini 2.0 Flash model. Speaker identities were clustered using ECAPA-TDNN embeddings.
Use Cases
Automatic Speech Recognition (ASR) for Uzbek
Text-to-Speech (TTS) synthesis for Uzbek
Fine-tuning speech models on Uzbek language data (e.g., Qwen3-TTS)
Speaker-conditioned TTS… See the full description on the dataset page: https://huggingface.co/datasets/openbank-uz/youtube_transcriptions.YouTube-Cantonese
Cantonese Audio Dataset from YouTube
This dataset contains Cantonese audio segments and creator uploaded transcripts (likely higher quality) extracted from various YouTube channels, along with corresponding transcript metadata. The data is intended for training automatic speech recognition (ASR) models.
Data Source and Processing
The data was obtained through the following process:
Download: Audio (.m4a) and available Cantonese subtitles (.srt for zh-TW, zh-HK, zh-Hant)… See the full description on the dataset page: https://huggingface.co/datasets/OrcinusOrca/YouTube-Cantonese.omnievalkit-dataset
OmniEvalKit Evaluation Datasets
Evaluation datasets for OmniEvalKit,
a comprehensive evaluation framework for omni-modal (audio + video + image + text) models.
Overview
Total subsets: 65
Total samples: 315,264
Total size: 620.3 GB (Parquet with embedded audio/image/video)
Subsets with embedded video: 15
Subsets requiring external video download: 2
Usage
from datasets import load_dataset
ds = load_dataset("OmniEvalKit/omnievalkit-dataset", "aishell1_test")… See the full description on the dataset page: https://huggingface.co/datasets/OmniEvalKit/omnievalkit-dataset.ghana-one-voice
Ghana One Voice
Speech in 43 Ghanaian and West African languages, all converted into a
single consistent voice.
Every clip keeps its original transcript, so the dataset pairs one speaker's
voice with the phonetic range of dozens of languages. Roughly 5 hours per
language, about 215 hours in total.
What this is
Source audio comes from many different speakers, recording conditions and
microphones. Each clip has been passed through
ghana-vc, a voice-conversion model… See the full description on the dataset page: https://huggingface.co/datasets/ghananlpcommunity/ghana-one-voice.