Skip to content
Fan Zeng (Frank)

Research resources

Speech & Audio Datasets

Frank curates this directory of publicly available speech and audio datasets. Access is free and requires no registration or application. Open a dataset’s details and select “Visit official source” for release information and access.

110 datasetsLast reviewed

110 datasets

  • GigaSpeechBench — Older Adults

    SpeechColab

    This benchmark provides Chinese and English older-adult speech with segment-level reference transcripts.

    • Aging and speech
    • Speech recognition

    Mandarin, English2 test configurations

    View details
  • Cambodian Oral History Project ASR

    Brigham Young University Matrix Lab and Cambodian Oral History Project

    Khmer oral-history interviews pair natural long-form speech with transcripts collected by the Cambodian Oral History Project.

    • Context, accents and devices
    • Speech recognition
    • Spoken interaction

    Khmer640 interviews, 356.6 hours of interview audio and 341.4 hours of aligned segments

    View details
  • LHCP-ASR Segments

    MLLP research group at Universitat Politècnica de València

    English talks from Large Hadron Collider Physics conferences pair recorded presentations with reference transcripts.

    • Context, accents and devices
    • Speech recognition

    English13,755 segments across two test splits

    View details
  • Hypa-Voices ASR Benchmark

    HypaAI and AfroVoices

    This compact evaluation set contains verbatim reference text for recordings across widely spoken and low-resource languages.

    • Context, accents and devices
    • Speech recognition

    English, French, Spanish +8200 audio and transcript pairs

    View details
  • GigaSpeechBench — Low-Resource Languages

    SpeechColab

    This module benchmarks language and dialect coverage, including Southeast Asian languages and regional Arabic varieties.

    • Context, accents and devices
    • Speech recognition

    Arabic, Indonesian, Malay +514 test configurations across language and regional varieties

    View details
  • GigaSpeechBench — Medical

    SpeechColab

    The Chinese and English medical subsets support recognition evaluation on domain-specific speech and terminology.

    • Clinical speech and vocabulary
    • Speech recognition

    Mandarin, English2 medical-domain test configurations

    View details
  • DuplexConv

    QualiaLabs and ASLP Lab

    DuplexConv provides large-scale Mandarin tutoring and casual conversation with model-assisted annotations of interaction events.

    • Spoken interaction
    • Meetings and diarization

    Mandarin2,000.21 hours across 93,709 files

    View details
  • ISCSLP 2026 CoT-TTS

    HKUST Audio

    This challenge corpus is designed for expressive synthesis that draws on surrounding dialogue and context annotations.

    • Speech synthesis

    Mandarin, EnglishAbout 16,000 hours across 3 million segments

    View details
  • SmoothConv

    QualiaLabs and ASLP Lab

    SmoothConv pairs naturally occurring Mandarin dialogue with expert annotations of turn events and conversational context.

    • Spoken interaction
    • Meetings and diarization

    Mandarin100.53 hours across 2,503 files

    View details
  • Medical Terms 2025

    Trelis Research

    This small synthetic benchmark pairs generated speech with labeled drug, condition and procedure terms.

    • Clinical speech and vocabulary
    • Speech recognition

    English50 synthetic test utterances with labeled target terms

    View details
  • Finnish Parliament ASR Benchmark

    Aalto University corpus, packaged by the GAIK project

    This release packages Finnish Parliament recordings for recognition evaluation, including subsets with speakers seen and unseen during original training.

    • Context, accents and devices
    • Speech recognition
    • Spoken interaction

    Finnish6,307 WAV and transcript pairs across five evaluation subsets

    View details
  • WAXAL ASR

    Google Research and WAXAL collection partners

    WAXAL brings together natural speech recordings from African language communities. Its labeled ASR splits pair audio with same-language text.

    • Context, accents and devices
    • Speech recognition

    Acholi, Akan, Amharic +16About 1,250 hours of transcribed speech across 19 languages

    View details
  • Omnilingual ASR Corpus

    Meta FAIR

    This Meta corpus contains transcribed spontaneous speech collected for languages that have limited representation in ASR research.

    • Speech recognition

    Multilingual3,350 hours across 348 languages

    View details
  • Taiwan Tongues ASR CE — Hakka

    Taiwan Administration for Digital Industries

    The Hakka release pairs speech files with reference sentences and duration metadata for each recording.

    • Context, accents and devices
    • Speech recognition

    Hakka2 train archives and one test archive

    View details
  • Taiwan Tongues ASR CE — Mandarin

    Taiwan Administration for Digital Industries

    Taiwanese Mandarin recordings are provided with Traditional Chinese references and locale metadata.

    • Context, accents and devices
    • Speech recognition

    Mandarin24 train archives and one test archive

    View details
  • Taiwan Tongues ASR CE — Taiwanese Hokkien

    Taiwan Administration for Digital Industries

    The Taiwanese Hokkien release provides recordings with corresponding written sentence references.

    • Context, accents and devices
    • Speech recognition

    Taiwanese Hokkien7 train archives and one test archive

    View details
  • Easy-Turn-Trainset

    ASLP Lab and Easy Turn contributors

    Easy-Turn-Trainset combines real and simulated interactions for learning when conversational systems should speak or listen.

    • Spoken interaction
    • Audio analysis

    Mandarin, EnglishAbout 1,145 hours of training data

    View details
  • LLaSO-Align

    EIT-NLP and LLaSO authors

    LLaSO-Align groups English audio and text pairs for the alignment phase of a speech-language model.

    • Speech recognition
    • Audio analysis

    EnglishAbout 12 million alignment samples

    View details
  • LLaSO-Instruct

    EIT-NLP and LLaSO authors

    LLaSO-Instruct organizes speech-language examples into instruction-following tasks for the LLaSO framework.

    • Spoken interaction
    • Audio analysis

    EnglishAbout 13.5 million instruction samples

    View details
  • Eka Medical ASR Evaluation Dataset

    EkaCare

    EkaCare combines spoken medical terms, narrated sentences and demonstration conversations in an Indian healthcare context.

    • Clinical speech and vocabulary
    • Speech recognition

    English, Hindi3,939 test clips (3,619 English, 320 Hindi)

    View details
  • ContextASR-Bench

    ContextASR-Bench research team

    This benchmark pairs generated English and Mandarin speech with contextual cues for named entities across multiple domains.

    • Context, accents and devices
    • Speech recognition

    English, Mandarin41,329 test entries across four language and task subsets

    View details
  • HiACC

    HiACC research team

    Adult and child speakers switch between Hindi and English while answering questions, reading stories and describing pictures.

    • Context, accents and devices
    • Speech recognition

    Hindi, English5.24 hours, 3,318 adult utterances and 1,858 child utterances

    View details
  • AISHELL-5

    AISHELL and AISHELL-5 research collaborators

    AISHELL-5 documents multi-speaker conversations in a car, with far-field door microphones and close-talk recordings.

    • Speech recognition
    • Meetings and diarization

    MandarinOver 100 hours of real multi-channel vehicle recordings

    View details
  • AfriSpeech-Dialog

    Intron Health and research collaborators

    This conversational benchmark pairs African-accented English audio with transcripts and diarization labels. Its healthcare scenarios sit alongside non-medical conversations.

    • Clinical speech and vocabulary
    • Speech recognition
    • Meetings and diarization
    • Clinical and accessible speech
    • Spoken interaction

    EnglishAbout 6 hours across 11 accents

    View details
  • DailyTalk Contiguous

    DailyTalk authors and Kyutai

    Two speakers record scripted English conversations derived from DailyDialog. The packaged audio places the speakers on separate stereo channels.

    • Speech recognition
    • Spoken interaction

    English2,541 recorded source dialogues

    View details
  • CS-Dialogue

    BAAI and Nankai University

    CS-Dialogue records spontaneous conversations that switch between Mandarin and English within the same interaction.

    • Speech recognition
    • Spoken interaction

    Mandarin, English104.02 hours

    View details
  • NOTSOFAR-1 Recorded Meetings

    Microsoft Research

    Far-field recordings of English meetings capture several talkers around a table using single-channel and multichannel devices.

    • Context, accents and devices
    • Speech recognition
    • Meetings and diarization

    English315 recorded meetings in the paper’s benchmark

    View details
  • MultiMed

    MultiMed research team

    MultiMed pairs medical-domain audio with transcriptions across five languages and multiple formats including interviews, talks and podcasts.

    • Clinical speech and vocabulary
    • Speech recognition

    Vietnamese, English, German +248,369 entries across published configuration splits

    View details
  • RealMAN

    Westlake University and AISHELL

    RealMAN provides real recordings made with a microphone array for speech enhancement and source localization.

    • Audio analysis

    Mandarin83.7 hours of speech signals and 144.5 hours of background noise

    View details
  • Speech-MASSIVE

    Fondazione Bruno Kessler and Speech-MASSIVE contributors

    Speakers record short voice-assistant requests with matching prompt text, intent and slot labels.

    • Speech recognition
    • Spoken interaction

    Arabic, German, Spanish +912 languages with 115 few-shot recordings per language

    View details
  • nEMO

    Adam Mickiewicz University CAI

    nEMO collects acted Polish speech labeled with emotional categories.

    • Audio analysis

    PolishOver 3 hours of annotated speech

    View details
  • VietMed

    Khai Le-Duc and collaborators

    VietMed combines labeled Vietnamese medical speech with unlabeled medical and general-domain audio.

    • Clinical speech and vocabulary
    • Speech recognition
    • Clinical and accessible speech

    Vietnamese16 labeled medical hours, 966 unlabeled medical hours and 1,204 unlabeled general-domain hours

    View details
  • EARS

    EARS research team

    EARS captures clean full-band speech in an anechoic chamber, including reading, freeform monologues and deliberate changes in delivery.

    • Aging and speech
    • Audio analysis
    • Speech recognition

    English100 hours from 107 speakers

    View details
  • NTU ML2021 ASR and Speech Translation

    National Taiwan University research team

    Recordings from a National Taiwan University machine learning course pair Mandarin and English lecture speech with text. The corpus includes spoken switches into English technical terms.

    • Context, accents and devices
    • Speech recognition

    Mandarin, English35,692 segments across train, development and test

    View details
  • Libriheavy

    Libriheavy authors and k2-fsa contributors

    Libriheavy aligns Libri-Light audiobook audio with English text while retaining punctuation, capitalization and surrounding context.

    • Speech recognition

    EnglishAbout 50,000 hours

    View details
  • RescueSpeech ASR

    RescueSpeech research team

    German speech from simulated search-and-rescue exercises is available with matched noise-augmented versions.

    • Context, accents and devices
    • Speech recognition

    German2,412 utterances, 1 h 36 min of reference speech

    View details
  • Corpus of Regional African American Language (CORAAL)

    Online Resources for African American Language · University of Oregon

    CORAAL pairs conversational recordings from several African American English communities with time-aligned orthographic transcriptions.

    • Context, accents and devices
    • Speech recognition
    • Spoken interaction

    EnglishOver 220 sociolinguistic interviews

    View details
  • LibriTTS-R

    Google

    LibriTTS-R applies learned audio restoration to the corresponding recordings from LibriTTS.

    • Speech synthesis

    EnglishAbout 585 hours

    View details
  • EdAcc

    University of Edinburgh CSTR

    EdAcc records English video-call conversations between people who know each other, with diverse accent and language backgrounds.

    • Context, accents and devices
    • Speech recognition

    EnglishApproximately 40 hours of conversations from 120 speakers

    View details
  • Expresso — Read Speech

    Meta AI Research

    Studio recordings cover expressive English reading styles including speech delivered softly, rapidly or with laughter.

    • Context, accents and devices
    • Speech recognition

    English11,615 rows in the published read configuration

    View details
  • The People's Speech

    MLCommons

    The People’s Speech assembles transcribed English from sources selected for reuse-friendly licensing.

    • Speech recognition

    EnglishOver 30,000 hours

    View details
  • Thai Elderly Speech

    VISAI and Data Wow

    Older Thai speakers read sentences about health symptoms and smart-home tasks.

    • Aging and speech
    • Speech recognition
    • Clinical and accessible speech

    Thai17 h 11 min, 19,200 recordings from 32 speakers

    View details
  • FLEURS

    Google

    FLEURS records translated sentences across many languages to support comparable multilingual speech evaluation.

    • Speech recognition
    • Audio analysis

    Multilingual, English, Mandarin +8102 languages with about 12 hours per language

    View details
  • NST Norwegian Speech Corpus

    Nordisk Språkteknologi and the National Library of Norway

    The Norwegian dictation corpus includes controlled recordings made with close and distant microphones.

    • Context, accents and devices
    • Speech recognition

    Norwegian9 training and 3 test shards per microphone configuration

    View details
  • PriMock57

    Babylon Health research team

    Clinicians and actors recorded simulated primary care visits with speaker-specific audio and manually annotated transcripts.

    • Clinical speech and vocabulary
    • Speech recognition
    • Clinical and accessible speech
    • Spoken interaction

    English57 consultations, 8 hours 38 minutes

    View details
  • TAU Urban Acoustic Scenes 2022 Mobile

    Tampere University

    This DCASE development set records urban acoustic scenes with reference and mobile devices.

    • Context, accents and devices
    • Audio analysis

    Language unconfirmed64 hours across 10 acoustic scenes

    View details
  • ASCEND

    HKUST Centre for Artificial Intelligence Research

    Bilingual speakers in Hong Kong hold unscripted conversations across multiple turns. The released segments include Mandarin speech, English speech and switches between the two.

    • Context, accents and devices
    • Speech recognition
    • Spoken interaction

    Mandarin, English10.62 hours, 12,314 utterances from 23 bilingual speakers

    View details
  • Earnings-22

    Rev.com research team

    Earnings-22 pairs long-form corporate earnings calls from multiple countries with verbatim reference transcripts.

    • Context, accents and devices
    • Speech recognition

    EnglishAbout 119 hours of earnings calls

    View details
  • MagicData-RAMC

    Magic Data Technology

    MagicData-RAMC pairs mobile recordings of Mandarin conversations with transcripts and topic labels.

    • Speech recognition
    • Meetings and diarization

    Mandarin180 hours

    View details
  • Norwegian Parliamentary Speech Corpus

    Norwegian Language Bank, National Library of Norway

    Recorded debates from the Norwegian Parliament pair public speech with time-aligned Bokmål and Nynorsk transcriptions.

    • Speech recognition
    • Spoken interaction

    Norwegian Bokmål, Norwegian Nynorsk125.7 hours excluding pauses, 140.3 hours including pauses

    View details
  • AliMeeting

    Alibaba DAMO Academy

    AliMeeting combines meeting-room microphone arrays and headset recordings with transcripts for multi-party Mandarin speech.

    • Speech recognition
    • Meetings and diarization

    Mandarin118.75 hours

    View details
  • ATCO2 Test Set 1h

    ATCO2 consortium

    Airline pilots and air traffic controllers communicate over live radio using short turns and domain-specific phraseology.

    • Context, accents and devices
    • Speech recognition

    EnglishAbout 1 hour, 871 segments in the data-card version

    View details
  • Kathbath

    AI4Bharat

    Contributors in India read text in 12 languages for a multilingual speech benchmark.

    • Speech recognition

    Bengali, Gujarati, Hindi +91,684 hours from 1,218 contributors

    View details
  • Kazakh Speech Corpus 2

    Institute of Smart Systems and Artificial Intelligence, Nazarbayev University

    KSC2 expands earlier Kazakh corpora with recordings from radio, television, parliamentary proceedings and podcasts.

    • Speech recognition

    KazakhAbout 1,200 hours and more than 600,000 utterances

    View details
  • Samrómur Children 21.09

    Reykjavik University Language and Voice Lab and Almannarómur

    Children and teenagers read Icelandic prompts for a crowdsourced speech collection.

    • Context, accents and devices
    • Speech recognition

    IcelandicMore than 137,000 validated recordings, ages 4–17

    View details
  • AISHELL-4

    Beijing AISHELL Technology

    AISHELL-4 records Mandarin meetings with an eight-channel circular microphone array.

    • Speech recognition
    • Meetings and diarization

    MandarinAbout 120 hours

    View details
  • Hi-Fi TTS

    NVIDIA

    Hi-Fi TTS selects high-quality English audiobook recordings from a small number of speakers for synthesis research.

    • Speech synthesis

    English291.6 hours from 10 speakers

    View details
  • VoxPopuli

    Facebook AI Research

    The transcribed VoxPopuli release pairs European Parliament speech with transcripts across several languages.

    • Speech recognition
    • Audio analysis

    Multilingual, English, German +41,791 transcribed hours in the main language table

    View details
  • Earnings-21

    Rev

    Public corporate earnings calls cover multiple industries and contain names, financial figures and dates.

    • Speech recognition
    • Spoken interaction

    English44 earnings calls, totaling 39 h 15 min

    View details
  • EasyCom

    Meta Reality Labs

    EasyCom captures group conversation around a table with wearable microphones, headset audio and synchronized egocentric video.

    • Context, accents and devices
    • Speech recognition
    • Meetings and diarization
    • Audio analysis

    English6.6 hours total, including about 5 h 18 min of quality-screened data

    View details
  • Golos

    SberDevices

    Russian speech was collected through crowdsourcing and recordings for far-field device use.

    • Context, accents and devices
    • Speech recognition

    RussianAbout 1,240 hours across crowd and far-field domains

    View details
  • MediaSpeech

    Nanosemantics Lab

    MediaSpeech pairs short clips from online media with transcripts in Arabic, French, Spanish and Turkish.

    • Speech recognition

    Arabic, French, Spanish +140 hours across 4 languages

    View details
  • MUCS 2021 Code-Switching ASR

    MUCS 2021 challenge organizers

    The challenge corpora capture English terms and phrases embedded in Hindi and Bengali technical tutorials.

    • Context, accents and devices
    • Speech recognition

    Hindi, Bengali, EnglishHindi–English 89.86 h train and 5.18 h test. Bengali–English 46.11 h train and 7.02 h test.

    View details
  • Multilingual TEDx

    Multilingual TEDx research team

    Multilingual TEDx aligns talk recordings with source-language transcripts and, where available, translated sentences.

    • Speech recognition

    Spanish, French, Portuguese +58 source languages, with language-specific archives

    View details
  • speechocean762

    Speechocean and corpus authors

    Mandarin-first-language speakers read English sentences. Pronunciation judgments cover sounds, words and complete utterances.

    • Context, accents and devices
    • Speech recognition

    English5,000 sentences, with child and adult speakers

    View details
  • West African Radio Corpus

    Nicolingua research team

    The corpus samples Guinean radio broadcasts with local languages, background music and telephone contributions.

    • Context, accents and devices
    • Audio analysis

    French, Guerze, Koniaka +717,090 thirty-second clips from 6 Guinean radio stations

    View details
  • West African Virtual Assistant Speech Recognition Corpus

    Moussa Doumbouya, Lisa Einstein and Chris Piech

    Speakers record short assistant commands, names and digits in four languages using laptops and phones.

    • Context, accents and devices
    • Speech recognition
    • Spoken interaction

    French, Maninka, Pular +110,083 utterances from 49 speakers aged 5–76

    View details
  • Yoloxóchitl Mixtec

    Yoloxóchitl Mixtec ASR research team

    Field recordings of Yoloxóchitl Mixtec include natural speech and a transcription system that represents lexical tone.

    • Speech recognition

    Yoloxóchitl MixtecAbout 100 hours of expert transcripts, plus 8.36 hours in the trainee correction study

    View details
  • Multilingual LibriSpeech (MLS)

    Facebook AI Research

    Multilingual LibriSpeech builds parallel dataset structure across eight languages from LibriVox recordings.

    • Speech recognition
    • Speech synthesis

    English, German, Dutch +5About 50,500 hours

    View details
  • SLURP

    SLURP research team

    SLURP links spoken assistant requests to text and semantic annotations for intents and entities.

    • Context, accents and devices
    • Speech recognition
    • Spoken interaction

    English18 domains, with separate real and synthetic audio

    View details
  • AISHELL-3

    Beijing AISHELL Technology

    AISHELL-3 provides neutral, multi-speaker Mandarin recordings with character and pinyin transcriptions.

    • Speech synthesis

    MandarinAbout 85 hours from 218 speakers

    View details
  • CHiME-6

    CHiME Challenge organizers

    CHiME-6 provides unscripted home conversations with wearable binaural microphones and multiple room arrays. The release corrects array synchronization from CHiME-5.

    • Context, accents and devices
    • Speech recognition
    • Meetings and diarization

    English20 fully transcribed dinner parties

    View details
  • Crowdsourced Peruvian Spanish Speech

    Google

    Volunteer speakers recorded Peruvian Spanish sentences, which are distributed with utterance-level text.

    • Context, accents and devices
    • Speech recognition
    • Speech synthesis

    Spanish2 WAV archives totaling about 1.9 GB

    View details
  • FSD50K

    Freesound and Universitat Pompeu Fabra research team

    FSD50K provides human-labeled sound events from Freesound, with clips that may carry several event labels.

    • Audio analysis

    Language unconfirmed51,197 clips, 108.3 hours and 200 sound classes

    View details
  • LibriCSS

    LibriCSS research team

    LibriCSS records overlapping playback of LibriSpeech utterances through a seven-channel array in a meeting room.

    • Context, accents and devices
    • Speech recognition
    • Meetings and diarization

    English10 hours across 10 sessions

    View details
  • Libri-Light

    Facebook AI Research and collaborators

    Libri-Light provides English audiobook speech for recognition with little or no labeled training data.

    • Speech recognition

    EnglishOver 60,000 hours of unlabeled audio

    View details
  • VCTK Corpus

    University of Edinburgh CSTR

    VCTK records English speakers with varied accents reading common and speaker-specific material for speech technology research.

    • Speech synthesis
    • Speaker recognition

    English110 speakers, about 400 sentences per speaker (v0.92)

    View details
  • CN-Celeb

    Tsinghua University CSLT

    CN-Celeb provides Chinese speaker recordings drawn from multiple media genres and recording conditions.

    • Speaker recognition

    MandarinAbout 274 hours (original CN-Celeb1)

    View details
  • JVS

    University of Tokyo research team

    JVS supports multi-speaker Japanese synthesis with recordings in normal, whispered and falsetto styles.

    • Speech synthesis
    • Speaker recognition

    JapaneseAbout 30 hours

    View details
  • LibriTTS

    Google

    LibriTTS restructures audiobook recordings and text from LibriSpeech for multi-speaker English synthesis.

    • Speech synthesis

    EnglishAbout 585 hours

    View details
  • Large Bengali ASR Training Corpus

    Google

    This corpus provides Bengali speech recordings with transcripts and anonymized speaker identifiers.

    • Speech recognition

    BengaliApproximately 196,000 utterances

    View details
  • Large Javanese ASR Dataset

    Google, Reykjavik University and Universitas Gadjah Mada

    This crowdsourced speech collection provides short Javanese recordings with utterance and speaker identifiers.

    • Speech recognition

    JavaneseAbout 185,000 utterances

    View details
  • Large Sundanese ASR Dataset

    Google

    This crowdsourced speech collection provides short Sundanese recordings with utterance and speaker identifiers.

    • Speech recognition

    SundaneseAbout 220,000 utterances

    View details
  • RAVDESS

    SMART Lab, Toronto Metropolitan University

    Professional actors recorded controlled speech and song performances with labeled emotions and intensity levels.

    • Audio analysis

    English1,440 speech-only audio files, plus song and video archives

    View details
  • Speech Commands

    Google

    Speech Commands contains short recordings of keywords and other words for small-footprint recognition.

    • Context, accents and devices
    • Speech recognition
    • Spoken interaction

    English105,829 one-second recordings across 35 words

    View details
  • TED-LIUM Release 3

    LIUM at Le Mans University

    Recorded TED talks are distributed as English speech segments with aligned reference text.

    • Context, accents and devices
    • Speech recognition

    English452 hours of training speech in Release 3

    View details
  • Zeroth-Korean

    Zeroth Project

    The Zeroth Project releases transcribed Korean recordings together with a pronunciation lexicon and text tools.

    • Speech recognition

    Korean51.6 hours for training and 1.2 hours for testing

    View details
  • JSUT

    University of Tokyo research team

    A Japanese reading corpus designed around everyday character readings, with recordings by a single speaker.

    • Speech synthesis

    JapaneseAbout 10 hours

    View details
  • AISHELL-1

    Beijing AISHELL Technology

    AISHELL-1 pairs read Mandarin utterances with transcriptions and includes speakers from multiple accent regions.

    • Speech recognition

    MandarinOver 170 hours from 400 speakers

    View details
  • VoiceBank-DEMAND

    University of Edinburgh CSTR

    The Edinburgh release provides paired noisy and clean utterances, plus text for speech enhancement and noise-robust speech workflows.

    • Context, accents and devices
    • Audio analysis
    • Speech recognition

    English28- and 56-speaker training configurations, with a separate test set

    View details
  • Room Impulse Response and Noise Database

    OpenSLR SLR28 release team

    The release combines measured and simulated room impulse responses with directional and diffuse noise.

    • Context, accents and devices
    • Audio analysis

    Language unconfirmed1.3 GB archive of room responses and noise

    View details
  • THCHS-30

    Tsinghua University CSLT

    A compact collection of Mandarin read speech with aligned transcripts and companion resources for noise testing.

    • Speech recognition

    MandarinAbout 30 hours

    View details
  • ESC-50

    Karol J. Piczak

    ESC-50 organizes short environmental audio into classes spanning animals, everyday activities and urban sounds.

    • Audio analysis

    Language unconfirmed2,000 five-second clips across 50 classes

    View details
  • LibriSpeech

    LibriSpeech authors / OpenSLR

    LibriSpeech provides aligned English audiobook recordings and transcripts with standard training, validation and test splits.

    • Speech recognition

    EnglishAbout 1,000 hours

    View details
  • MUSAN

    David Snyder, Guoguo Chen and Daniel Povey

    MUSAN collects music, multilingual speech and a range of technical and everyday noise recordings.

    • Audio analysis

    Multilingual11 GB archive with music, speech and noise

    View details
  • Seoul Corpus

    Seoul Corpus research team

    The Seoul Corpus pairs spontaneous Korean speech with phonetic annotation.

    • Context, accents and devices
    • Speech recognition

    Korean2.5 GB of FLAC audio plus separate TextGrid annotations

    View details
  • UrbanSound8K

    New York University research team

    UrbanSound8K collects labeled excerpts of city sounds such as sirens, street music and road equipment.

    • Audio analysis

    Language unconfirmed8,732 clips across 10 sound classes

    View details
  • MATCH Corpus

    University of Edinburgh

    MATCH records younger and older adults as they interact with a Wizard-of-Oz spoken dialogue system.

    • Aging and speech
    • Speech recognition
    • Spoken interaction

    English24 older and 22 younger participants (project description)

    View details
  • ATCOSIM

    EUROCONTROL Experimental Centre and Graz University of Technology

    Controllers recorded through headsets during air traffic management simulations speak English in an aviation setting.

    • Context, accents and devices
    • Speech recognition

    EnglishAbout 10 hours, 10,078 utterances from 10 speakers

    View details
  • Santa Barbara Corpus of Spoken American English

    University of California, Santa Barbara

    Recordings of everyday American English include face-to-face conversations, telephone calls and storytelling.

    • Context, accents and devices
    • Speech recognition
    • Spoken interaction

    English60 recordings, typically about 20 minutes each

    View details
  • ICSI Meeting Corpus

    International Computer Science Institute

    ICSI provides natural research meetings with synchronized audio and time-aligned orthographic transcripts.

    • Context, accents and devices
    • Speech recognition
    • Meetings and diarization
    • Spoken interaction

    EnglishAbout 72 hours of speech across 75 meetings

    View details
  • AMI Meeting Corpus

    AMI Project Consortium

    AMI pairs synchronized meeting audio and annotations with close-talking and distant microphone recordings.

    • Context, accents and devices
    • Speech recognition
    • Meetings and diarization
    • Spoken interaction

    EnglishAbout 100 hours of meetings

    View details
  • CML-TTS

    Universidade Federal de Minas Gerais research team

    CML-TTS prepares multilingual LibriVox audiobook recordings for text-to-speech research in seven languages.

    • Speech synthesis

    Multilingual, Dutch, German +53,233.43 hours across seven languages

    View details
  • Crowdsourced Burmese Speech

    Google

    The public release pairs Burmese recordings from female speakers with a separate text index.

    • Speech recognition

    Burmese948 MB audio archive (publisher listing)

    View details
  • KSS

    Kyubyong Park

    KSS collects recordings from one Korean speaker for supervised text-to-speech research.

    • Speech synthesis

    Korean12,853 clips, over 12 hours

    View details
  • MagicData Mandarin Chinese Read Speech Corpus

    Magic Data Technology

    MagicData’s read-speech corpus contains scripted Mandarin recorded by speakers from mainland China.

    • Speech recognition

    Mandarin755 hours from 1,080 speakers

    View details
  • Primewords Chinese Corpus Set 1

    Shanghai Primewords

    Primewords Chinese Corpus Set 1 contains transcribed Mandarin utterances recorded with mobile devices.

    • Speech recognition

    MandarinAbout 100 hours

    View details
  • ST-CMDS (Free ST Chinese Mandarin Corpus)

    Surfingtech

    ST-CMDS provides short Mandarin utterances collected from hundreds of speakers for speech recognition.

    • Speech recognition

    Mandarin102,600 utterances from 855 speakers

    View details
  • The LJ Speech Dataset

    Keith Ito and Linda Johnson

    The LJ Speech Dataset segments one narrator’s readings from public-domain books into English audio and text pairs.

    • Speech synthesis

    English13,100 clips, about 24 hours

    View details
  • TORGO

    University of Toronto and research collaborators

    TORGO pairs acoustic and articulatory recordings of speakers with dysarthria and a comparison group.

    • Aging and speech
    • Speech recognition
    • Clinical and accessible speech

    English8 speakers with dysarthria and 7 controls

    View details