datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
genshin-voice
Genshin Voice
Genshin Voice is a dataset of voice lines from the popular game Genshin Impact.
Hugging Face 🤗 Genshin-Voice
ModelScope Genshin-Voice
Per-speaker downloads are grouped by language and ZIP size. Browse every archive in the ZIP index.
Last update at 2026-08-13
654252 wavs
7291 without speaker (1%)
52693 without transcription (8%)
1088 without inGameFilename (0%)
Dataset Details
Dataset Description
The dataset contains voice lines… See the full description on the dataset page: https://huggingface.co/datasets/simon3000/genshin-voice.coral-v3
CoRal: Danish Conversational and Read-aloud Dataset
Version 3.0
Dataset Overview
CoRal is a comprehensive Automatic Speech Recognition (ASR) dataset designed to capture the diversity of the Danish language across various dialects, accents, genders, and age groups. The primary goal of the CoRal dataset is to provide a robust resource for training and evaluating ASR models that can understand and transcribe spoken Danish in all its variations.
Key Features… See the full description on the dataset page: https://huggingface.co/datasets/CoRal-project/coral-v3.shrutilipizenless-voice
Zenless Voice
Zenless Voice is a dataset of voice lines from the popular game Zenless Zone Zero.
Hugging Face 🤗 Zenless-Voice
ModelScope Zenless-Voice
Per-speaker downloads are grouped by language and WAV count. Browse every archive in the ZIP index.
Last update at 2026-09-17, game version 3.2.0
406720 wavs
78785 without speaker (19%)
123429 without transcription (30%)
83509 without inGameFilename (21%)
Speaker archives contain 327,935 WAVs in 4,322 ZIPs. The 78,785 rows… See the full description on the dataset page: https://huggingface.co/datasets/simon3000/zenless-voice.genshin-voice-v3.3-mandarin
Dataset Card for Genshin Voice
Dataset Description
Dataset Summary
The Genshin Voice dataset is a text-to-voice dataset of different Genshin Impact characters unpacked from the game.
Languages
The text in the dataset is in Mandarin.
Dataset Creation
Source Data
Initial Data Collection and Normalization
The data was obtained by unpacking the Genshin Impact game.
Who are the source language producers?
The… See the full description on the dataset page: https://huggingface.co/datasets/hanamizuki-ai/genshin-voice-v3.3-mandarin.liepa-3
LIEPA-3 — Lithuanian Speech Corpus
Didysis lietuvių kalbos garsynas (LIEPA-3)
Dataset Summary
LIEPA-3 is a large, open corpus of Lithuanian speech (~10,000 hours,
~7.5 million audio files) built for automatic speech recognition (ASR),
text-to-speech (TTS) and linguistic research. It spans read, spontaneous,
phonetically-annotated and dialectal speech recorded under a wide range of
conditions (studio, dictaphone, radio, TV, telephone, audiobooks).
Official… See the full description on the dataset page: https://huggingface.co/datasets/meldynamics/liepa-3.starrail-voice
StarRail Voice
StarRail Voice is a dataset of voice lines from the popular game Honkai: Star Rail.
Hugging Face 🤗 StarRail-Voice
ModelScope StarRail-Voice
Last update at 2026-07-16, game version 4.4.0
403437 wavs
60164 without speaker (15%)
61375 without transcription (15%)
57869 without inGameFilename (14%)
Dataset Details
Dataset Description
The dataset contains voice lines from the game's characters in multiple languages, including Chinese… See the full description on the dataset page: https://huggingface.co/datasets/simon3000/starrail-voice.ISSAI_KSC_335RS_v_1_1
Dataset Card for "ISSAI_KSC_335RS_v_1_1"
Kazakh Speech Corpus (KSC)
Identifier: SLR102
Summary: A crowdsourced open-source Kazakh speech corpus developed by ISSAI (330 hours)
Category: Speech
License: Attribution 4.0 International (CC BY 4.0)
Downloads (use a mirror closer to you):
ISSAI_KSC_335RS_v1.1_flac.tar.gz [19G] (speech, transcripts and metadata ) Mirrors: [US] [EU] [CN]
About this resource:
A crowdsourced open-source speech corpus for the Kazakh language. The KSC… See the full description on the dataset page: https://huggingface.co/datasets/Shirali/ISSAI_KSC_335RS_v_1_1.warsh-segments-v3
Haitam03/warsh-v3
Warsh (Rewayat Warsh A'n Nafi') Quran recitation, segmented at waqf with
obadx/recitation-segmenter-v2.
Built with warsh-data.
Layout
path
what
data/<reciter>/<surah>.parquet
one file per source recording, audio embedded as 16 kHz mono FLAC
raw/<reciter>/<surah>.mp3
the source recording it came from
segment_params.json
the settings this corpus was produced with
One parquet per source recording, named after it, so re-running a… See the full description on the dataset page: https://huggingface.co/datasets/Haitam03/warsh-segments-v3.GLOBE_V3
Important notice
Differences between V3 version and two previous versions (V1|V2):
This version is built base on Common Voice 21.0 English Subset.
This version only includes utterance that are an exact match with the transcription from Whisper V3 LARGE (CER == 0).
This version includes the original Common Voice metadata (age, gender, accent, and ID).
All audio files in this version are at 24kHz sampling rate.
All audio files in this version are unenhanced. (We’d greatly… See the full description on the dataset page: https://huggingface.co/datasets/MushanW/GLOBE_V3.pseudolabel-malaysian-youtube-whisper-large-v3
Pseudolabel Malaysian Youtube videos using Whisper Large V3
Original dataset at https://huggingface.co/datasets/malaysia-ai/crawl-youtube, distributed pseudolabelled using 4x A100s
script at https://github.com/mesolitica/malaysian-dataset/tree/master/speech-to-text-semisupervised/pseudolabel-whisper
Each audio is 30 seconds.
Each audio saved in 16k sample rate.
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.genshin-voice-v3.5-mandarin
Dataset Card for Genshin Voice
Dataset Description
Dataset Summary
The Genshin Voice dataset is a text-to-voice dataset of different Genshin Impact characters unpacked from the game.
Languages
The text in the dataset is in Mandarin.
Dataset Creation
Source Data
Initial Data Collection and Normalization
The data was obtained by unpacking the Genshin Impact game.
Who are the source language producers?
The… See the full description on the dataset page: https://huggingface.co/datasets/hanamizuki-ai/genshin-voice-v3.5-mandarin.seamless-interaction-jefferson-annotations
Seamless Interaction Jefferson-Style Annotations
An automatic, turn-oriented annotation layer for the
Meta Seamless Interaction Dataset.
It compares the dataset's traditional transcript with an ASR-derived
Jefferson-style condition and supplies speech-act, communicative-purpose,
interactional-signal, alignment, and quality fields.
This is a derived noncommercial research dataset. It does not redistribute
the source audio. Every record retains the original interaction ID, split… See the full description on the dataset page: https://huggingface.co/datasets/kennethli319/seamless-interaction-jefferson-annotations.afri-temp-data3
AfricanVoices Hausa -- Train Split
from datasets import load_dataset
ds = load_dataset("suleiman2003/afri-temp-data3", split="train")
print(ds[0])
liepa3
LIEPA-3 Lithuanian Speech Corpus
This repository repackages the original LIEPA-3 release into Hugging Face
Parquet shards with embedded FLAC audio bytes. The original transcriptions are
kept as released: normalized lowercase Lithuanian text without punctuation,
digits, capitalization, or other symbols.
Recommended use:
read: cleanest subset and the default starting point for TTS or ASR.
spon: spontaneous/broadcast/media speech; useful for ASR, not a clean TTS default.
dial:… See the full description on the dataset page: https://huggingface.co/datasets/i4tech/liepa3.ivirits-audio-v2-30s
ivrit.ai audio-v2 — 2–30 s segments
ivrit-ai/audio-v2 (>20k hours of Hebrew
audio) cut into 2–30 second speech segments with machine transcripts, ready for ASR
fine-tuning.
How it was built
VAD — Silero VAD (ONNX) over each episode decoded to 16 kHz mono. Speech regions
longer than 30 s are split at the quietest sufficiently-long pause inside the window,
so cuts land in silence rather than mid-word. Regions shorter than 2 s are dropped.
Transcription —… See the full description on the dataset page: https://huggingface.co/datasets/notmax123/ivirits-audio-v2-30s.nasle-mana-clean-chunked-30s
Nasl-e-Mana Clean Speech Corpus — Sentence-Safe 30s Chunks
Training-oriented WAV chunks derived from the public Nasl-e-Mana magazine audio corpus. Chunks target approximately 30 seconds and are cut at detected acoustic pauses; the labeled configuration additionally assigns only complete source-text sentences to each chunk.
Configuration
Rows
Columns
Meaning
labeled (train/)
9,886
audio, label
Sentence-grouped text/audio pairs from duration-compatible source-text… See the full description on the dataset page: https://huggingface.co/datasets/Reza2kn/nasle-mana-clean-chunked-30s.nasle-mana-clean-chunked-30s-avasanj
Nasl-e-Mana Clean Persian Speech — corrected 30-second chunks
Corrected, provenance-preserving audio chunks collected from the Nasl-e-Mana magazine website, generated on 2026-08-30. This release supersedes the earlier unreliable proportional-mapping chunk export; that older release was not used here.
Splits
Split
Rows
Audio
Columns
labeled
4,981
41.41 hours
audio, label
to_transcribe
11,127
92.72 hours
audio
The labeled split contains the… See the full description on the dataset page: https://huggingface.co/datasets/Reza2kn/nasle-mana-clean-chunked-30s-avasanj.global-news-radio-30s
Global News Radio Dataset
Multilingual news radio recordings from 51 languages across 42 countries.
Recordings
51
Total audio
1500 min (25.0 h)
Format
MP3 16kHz mono 64kbps
Parquet shards
11
Languages
51
Countries
42
Size
687 MB
Languages
Amharic, Arabic, Bashkir, Basque, Belarusian, Bengali, Brazilian Portuguese,Portugues Do Brasil,Português Brasil, Catalan, Croatian, Czech, Danish, Dutch, English, Estonian, Faroese, Finnish, Flemish… See the full description on the dataset page: https://huggingface.co/datasets/NathanRoll/global-news-radio-30s.Meta_STT_ZH_AIShell3
Meta Speech Recognition Mandarin Dataset (AISHELL3)
This dataset contains both metadata and audio files for Mandarin speech recognition samples from the AISHELL3 corpus.
Dataset Statistics
Splits and Sample Counts
train: 60098 samples
valid: 3163 samples
test: 24772 samples
Example Samples
train
{
"audio_filepath": "/external4/datasets/Mandarin/AISHELL3/wavs_train/SSB00430356.wav",
"text": "她以 ENTITY_PRODUCT 滴鸡精 END 调养身体。 AGE_14_25… See the full description on the dataset page: https://huggingface.co/datasets/WhissleAI/Meta_STT_ZH_AIShell3.parlament_parla_v3
Dataset Card for ParlamentParla v3 - Speech Corpus of Catalan Parliamentary Sessions
A speech corpus composed of Catalan Parliamentary Sessions.The v3 and last version of the corpus includes both clean and other quality segments, divided into short segments (less than 30 seconds) and long segments (more than 30 seconds). The total dataset encompasses 1059h 48m 04s of speech, including 945h 51m 06s for the short segments and 113h 56m 58s for the long segments, with a total of… See the full description on the dataset page: https://huggingface.co/datasets/projecte-aina/parlament_parla_v3.WaxalNLP
Waxal Datasets
The WAXAL dataset is a large-scale multilingual speech corpus for African languages, introduced in the paper WAXAL: A Large-Scale Multilingual African Language Speech Corpus.
Dataset Description
The Waxal project provides datasets for both Automated Speech Recognition (ASR)
and Text-to-Speech (TTS) for African languages. The goal of this dataset's
creation and release is to facilitate research that improves the accuracy and
fluency of speech and language… See the full description on the dataset page: https://huggingface.co/datasets/novelwolde36/WaxalNLP.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.genshin-voice-v3.4-mandarin
Dataset Card for Genshin Voice
Dataset Description
Dataset Summary
The Genshin Voice dataset is a text-to-voice dataset of different Genshin Impact characters unpacked from the game.
Languages
The text in the dataset is in Mandarin.
Dataset Creation
Source Data
Initial Data Collection and Normalization
The data was obtained by unpacking the Genshin Impact game.
Who are the source language producers?
The… See the full description on the dataset page: https://huggingface.co/datasets/hanamizuki-ai/genshin-voice-v3.4-mandarin.Egyptian-ASR-MGB-3
Egyptian Arabic dialect automatic speech recognition
Dataset Summary
This dataset was collected, cleaned and adjusted for huggingface hub and ready to be used for whisper finetunning/training.
From MGB-3 website:
The MGB-3 is using 16 hours multi-genre data collected from different YouTube channels. The 16 hours have been manually transcribed.
The chosen Arabic dialect for this year is Egyptian.
Given that dialectal Arabic has no orthographic rules, each program has… See the full description on the dataset page: https://huggingface.co/datasets/MightyStudent/Egyptian-ASR-MGB-3.thchs30W_hausa_v3
Cleaned Hausa Speech Dataset v3
A cleaned and processed Hausa speech dataset built from multiple open-source Hugging Face datasets.
Dataset Description
This dataset contains cleaned, normalized, and deduplicated Hausa speech audio with aligned transcriptions. All audio is:
Sample rate: 16,000 Hz (mono)
Format: FLAC (lossless, embedded in Parquet)
Duration range: 1–30 seconds per clip
Loudness normalized: -20 dBFS RMS
VAD trimmed: Non-speech segments removed with… See the full description on the dataset page: https://huggingface.co/datasets/suleiman2003/W_hausa_v3.asr-farsi-youtube-chunked-30-seconds
How To Use
from datasets import load_dataset
train = load_dataset('pourmand1376/asr-farsi-youtube-chunked-30-seconds', split='train+val')
test =load_dataset('pourmand1376/asr-farsi-youtube-chunked-30-seconds', split='test')
+300 Hours ASR dataset generated from this kaggle dataset
MGB-3-Arabic
Dataset Card for MGB-3 Arabic Speech Recognition
Dataset Summary
The MGB-3 Arabic dataset is a multi-genre collection of Egyptian Arabic speech extracted from YouTube videos, designed for speech recognition in challenging, real-world conditions. Unlike its predecessor MGB-2 which focused on broadcast TV news, MGB-3 emphasizes dialectal Arabic across diverse content types.
The dataset contains approximately 16 hours of Egyptian Arabic speech from 80 YouTube videos… See the full description on the dataset page: https://huggingface.co/datasets/MohamedRashad/MGB-3-Arabic.