126 of 126Human speech and open licences first
| Save | FD | Length | Domain | Languages | Licence | Source | Speech | |||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| — | 42.6k h29.3k validated | — | — | — | — | 295 | Permissive | Consented | human | 2019-06 | ||
| — | ~3,500 hconv. | — | — | — | — | hi bn ta +19 | Permissive | Consented | human | 2024-03 | ||
| — | 2,880 h | — | — | — | — | 21 + en | Permissive | Consented | human | 2020-07 | ||
| — | ~1k h | — | — | — | — | 102+ | Permissive | Consented | human | 2022-05 | ||
| Full-duplex | 280 h | — | — | Casual talk | — | en | Permissive | Consented | human | 2026-02 | ||
AISHELL-1Beijing Shell Shell Technology | — | 178 h | — | — | — | — | zh | Permissive | Consented | human | 2017-09 | |
| Full-duplex | 141 h | — | — | Casual talk | — | en | Permissive | Consented | human | 2026-02 | ||
| — | 100 h | — | — | — | — | en | Permissive | Consented | human | 2006-06 | ||
AISHELL-3Beijing Shell Shell Technology | — | 85 h | — | — | — | — | zh | Permissive | Consented | human | 2020-10 | |
ICSI Meeting CorpusICSI / University of Edinburgh (AMI site) | — | 72 h | — | — | — | — | en | Permissive | Consented | human | 2004-02 | |
VCTK CorpusCSTR, University of Edinburgh | — | 44 h | — | — | — | — | en | Permissive | Consented | human | 2019-11 | |
| Full-duplex | 23.7 h | 59 | — | Customer service | — | en | Permissive | Consented | human | 2020-10 | ||
| Full-duplex | 20 h | — | avg ~21 min | Task-oriented | — | en | Permissive | Consented | human | 2026-08 | ||
| Full-duplex | 18 h | 64 | — | Task-oriented | — | en | Permissive | Consented | human | — | ||
| Full-duplex | 17.4 h | — | avg ~22 min | Task-oriented | — | en | Permissive | Consented | human | 2026-09 | ||
| — | ~9 h | — | — | — | — | en | Permissive | Consented | human | 2022-05 | ||
| — | 5.3 h | — | — | — | — | en | Permissive | Consented | human | 2019-09 | ||
| Full-duplex | 5.1 h | 8 | avg 6.8 min | Customer service | — | en | Permissive | Consented | human | 2026-10 | ||
| — | ~108 h | — | — | — | — | hi bn ta +19 | Permissive | Mixed | human | 2026-07 | ||
| — | 1.1M h | — | — | — | — | 100+ | Permissive | Public media | human | 2026-09 | ||
YODAS / YODAS2WavLab (CMU) | — | 500k h | — | — | — | — | 100+ | Permissive | Public media | human | 2024-06 | |
| — | 400k h | — | — | — | — | en de fr +20 | Permissive | Public media | human | 2021-01 | ||
| — | 60k h | — | — | — | — | en | Permissive | Public media | human | 2019-12 | ||
| — | 50k h | — | — | — | — | en de nl +5 | Permissive | Public media | human | 2020-12 | ||
| — | 6,628 h | — | — | — | — | multi | Permissive | Public media | human | 2020-11 | ||
LibriSpeechVassil Panayotov et al. | — | 960 h | — | — | — | — | en | Permissive | Public media | human | 2015-04 | |
| — | 585 h | — | — | — | — | en | Permissive | Public media | human | 2019-04 | ||
| — | 585 h | — | — | — | — | en | Permissive | Public media | human | 2023-05 | ||
| — | 292 h | — | — | — | — | en | Permissive | Public media | human | 2021-04 | ||
CSS10Kyubyong Park et al. | — | 140 h | — | — | — | — | de el es +7 | Permissive | Public media | human | 2019-03 | |
MUSANJohns Hopkins University | — | 109 h | — | — | — | — | multi | Permissive | Public media | human | 2015-10 | |
LJSpeechKeith Ito | — | 24 h | — | — | — | — | en | Permissive | Public media | human | 2017-07 | |
Real-TurnTurkAta Technology Platforms | Full-duplex | 1.94 hpublic subset | 6 | avg ~19 min | — | — | tr | Permissive | Not stated | human | 2026-08 | |
| Full-duplex | 128.6 h | 156 | 5–15 min | Customer service | — | en | Share-alike | Consented | human | 2026-04 | ||
| — | 118.75 h | — | — | — | — | zh | Share-alike | Consented | human | — | ||
CREMA-DUniversity of Pennsylvania | — | ~5 h | — | — | — | — | en | Share-alike | Consented | human | 2014-08 | |
| — | 30k+ h | — | — | — | — | en | Share-alike | Public media | human | 2021-11 | ||
| — | 119 h | — | — | — | — | en | Share-alike | Public media | human | 2022-03 | ||
| — | 20 h | — | — | — | — | en | Share-alike | Consented | human | 2022-07 | ||
| — | 643k h192k h with audio | — | — | — | — | en de fr +22 | Permissive | Public media | human | 2025-05 | ||
| — | 35k h | — | — | — | — | ja | Share-alike | Public media | human | 2023-03 | ||
VoxCeleb2VGG, University of Oxford | — | 2,442 h | — | — | — | — | multi | Share-alike | Public media | human | 2018-06 | |
VoxCeleb1VGG, University of Oxford | — | 352 h | — | — | — | — | en | Share-alike | Public media | human | 2017-06 | |
| Full-duplex | 4,000+ h | 4,284 | 2–10 min | Casual talk, Task-oriented, Other | — | en | Non-commercial | Consented | human | 2025-06 | ||
| Full-duplex | 2,100 h100 h + 2,000 h | — | clips avg 1.3–2.4 min | Casual talk, Other | — | zh | Non-commercial | Consented | human | 2026-06 | ||
MyST Children's Conversational SpeechBoulder Learning / LDC | — | 393 h | — | — | — | — | en | Non-commercial | Consented | human | 2021-06 | |
MultiDialogIVLLab (Yonsei / Naver) | — | 340 h | — | — | — | — | en | Non-commercial | Consented | human | 2024-06 | |
| — | 249 h | — | — | — | — | en | Non-commercial | Consented | human | 2023-05 | ||
| — | 180 h | — | — | — | — | zh | Non-commercial | Consented | human | 2022-03 | ||
AISHELL-5Beijing AISHELL Technology / OpenSLR SLR-159 | — | 105 h | — | — | — | — | zh | Non-commercial | Consented | human | 2025-05 | |
CASPERJohns Hopkins University (The SSSD Project) | Full-duplex | 102 h | 157 | avg 15.8 min | Casual talk | — | en | Non-commercial | Consented | human | 2025-06 | |
WHAM! / WHAMR!Whisper.ai / MERL | — | ~80 h | — | — | — | — | en | Non-commercial | Consented | human | 2019-07 | |
CHiME-6CHiME Challenge | — | 40 h | — | — | — | — | en | Non-commercial | Consented | human | 2020-04 | |
| — | 40 h | — | — | — | — | en | Non-commercial | Consented | human | 2023-08 | ||
| Full-duplex | 15 h10 zh + 5 en | 14 | avg ~26 min | Casual talk | — | zh en | Non-commercial | Consented | human | 2024-11 | ||
| Full-duplex | 10.35 h | — | — | — | — | ja | Non-commercial | Consented | human | 2025-04 | ||
RAVDESSRyerson University | — | ~3 h | — | — | — | — | en | Non-commercial | Consented | human | 2018-04 | |
| — | 215.6k hEmilia-Large | — | — | — | — | en zh ja +3 | Non-commercial | Public media | human | 2024-07 | ||
J-CHATSarulab (UTokyo) / Keio | — | 76k h | — | — | — | — | ja | Non-commercial | Public media | human | 2024-07 | |
| — | 10k hlabelled | — | — | — | — | zh | Non-commercial | Public media | human | 2021-10 | ||
MuST-CFBK (Fondazione Bruno Kessler) | — | ~4k h | — | — | — | — | en → 14 | Non-commercial | Public media | human | 2019-06 | |
Multilingual TEDxJohns Hopkins / CMU | — | 760 h | — | — | — | — | es fr pt +5 | Non-commercial | Public media | human | 2021-02 | |
| — | 452 h | — | — | — | — | en | Non-commercial | Public media | human | 2018-05 | ||
CN-CelebCSLT, Tsinghua University | — | 274 h | — | — | — | — | zh | Non-commercial | Public media | human | 2019-11 | |
| — | 100 hopen subset | — | — | — | — | en | Non-commercial | Public media | human | 2025-09 | ||
| Full-duplex | 1,960 h | 11,917 | up to 10 min | Casual talk | $42,000$21/h | en | Gated | Consented | human | 2004-12 | ||
| — | 1,604 h | — | — | — | — | en fr de +8 | Gated | Consented | human | 2025-09 | ||
| Full-duplex | 1,502 h | 2,645 | median 33 min | Casual talk | — | en es fr +19 | Gated | Consented | human | 2026-09 | ||
| Full-duplex | 1,000 h8.9 h open sample | 239 | avg 37.5 min | Casual talk | — | en | Gated | Consented | human | 2026-09 | ||
AISHELL-2Beijing Shell Shell Technology | — | 1,000 h | — | — | — | — | zh | Gated | Consented | human | 2018-08 | |
KsponSpeechETRI / Korean AI Hub | — | 969 h | — | — | — | — | ko | Gated | Consented | human | 2020-03 | |
| Full-duplex | 727 h | 209 | avg 25–30 min | Casual talk | — | en | Gated | Consented | human | 2025-06 | ||
| — | 661 h | — | — | — | — | ja | Gated | Consented | human | 2004-01 | ||
| Full-duplex | 590 h | — | — | Casual talk | $37,750$64/h | en | Gated | Consented | human | 2013-06 | ||
| Full-duplex | ~313 h10 sets with stated hours | — | 5–30 min | Casual talk | $39,750$127/h | en fr ar +10 | Gated | Consented | human | — | ||
| Full-duplex | 260 h | 543 | avg ~6.5 min | Casual talk | $35,500$137/h | en | Gated | Consented | human | 1993-01 | ||
| Full-duplex | ~229 h2026 Second Editions | — | up to 30 min | Casual talk | — | ja zh es +3 | Gated | Consented | human | 1996-01 | ||
| Full-duplex | 149 h | 1,793 | up to 10 min | Casual talk | $36,000$242/h | zh | Gated | Consented | human | 2005-07 | ||
| Full-duplex | 104.9 h | — | avg 15 min | Casual talk, Task-oriented, Other | — | en | Gated | Consented | human | 2026-06 | ||
| — | ~80 h | — | — | — | — | en | Gated | Consented | human | 1993-01 | ||
| Full-duplex | 42.1 h | 682 | clips avg 2.4 min | Casual talk | — | en zh ja +18 | Gated | Consented | human | 2026-10 | ||
IEMOCAPUSC SAIL | — | 12 h | — | — | — | — | en | Gated | Consented | human | 2008-12 | |
TIMITDARPA / LDC | — | 5.4 h | — | — | — | $34,125$6,319/h | en | Gated | Consented | human | 1993-10 | |
GigaSpeechSpeechColab | — | 10k h | — | — | — | — | en | Gated | Public media | human | 2021-06 | |
| — | 5,000 h | — | — | — | — | en | Gated | Public media | human | 2021-04 | ||
| — | 409 hv2.0 | — | — | — | — | en | Gated | Public media | human | 2017-01 | ||
| Full-duplex | 1,500 h | 150 | — | — | $135,000$90/h | en | Custom | Consented | human | — | ||
| Full-duplex | 1,125 h | 565 | median 15.5 min | Casual talk | $11,250$10/h | en | Custom | Consented | human | 2026-11 | ||
| Full-duplex | ~860 h | 1,460 | median 29 min | Casual talk | — | en | Custom | Consented | human | 2022-02 | ||
JVS CorpusTakamichi Lab (University of Tokyo) | — | 30 h | — | — | — | — | ja | Custom | Consented | human | 2019-08 | |
ESD (Emotional Speech Dataset)NUS / HLT Lab | — | 29 h | — | — | — | — | en zh | Custom | Consented | human | 2021-05 | |
L2-ARCTICTexas A&M / Iowa State | — | 27 h | — | — | — | — | en | Custom | Consented | human | 2018-09 | |
Dialogs (Russian)Shigabeev, Latyshev | — | 20.6 h | — | — | — | — | ru | Custom | Consented | human | 2026-07 | |
JSUTShinnosuke Takamichi (U. Tokyo) | — | 10 h | — | — | — | — | ja | Custom | Consented | human | 2017-11 | |
| Full-duplex | — | — | — | Casual talk | — | en | Custom | Consented | human | — | ||
| Full-duplex | — | — | — | Casual talk | — | en | Custom | Consented | human | — | ||
| Full-duplex | — | — | — | Casual talk | — | en | Custom | Consented | human | — | ||
| Full-duplex | — | — | — | Casual talk | — | ko | Custom | Consented | human | — | ||
| Full-duplex | — | — | — | Casual talk | — | ja | Custom | Consented | human | — | ||
| Full-duplex | — | — | — | — | — | ja | Custom | Consented | human | — | ||
| Full-duplex | — | — | — | — | — | ja | Custom | Consented | human | — | ||
| Full-duplex | — | — | — | Task-oriented | — | en | Custom | Consented | human | — | ||
DuplexChatNakata, Saito, Saruwatari (U. Tokyo / AIST) | Full-duplex | 415k h | — | clips avg 1.1 min | Podcasts and media | — | en ja | Custom | Public media | human | 2026-07 | |
MM-F2FXiamen University | — | 210 h | — | — | — | — | en | Custom | Public media | human | 2025-07 | |
| — | ~12 h | — | — | — | — | en | Custom | Public media | human | 2018-10 | ||
Global Conversational SpeechUsergyAI | Full-duplex | 305 h | — | 10–60 min | Healthcare, Meetings, Customer service | $19,825$65/h | zh en es +8 | Custom | Not stated | human | 2026-04 | |
| — | 3,809 h | — | — | — | — | 21 → en | Permissive | Consented | mixed | 2022-01 | ||
NaijaS2STMaltais et al. | — | ~200 h | — | — | — | — | ig ha yo +2 | Permissive | Consented | mixed | 2026-04 | |
| — | — | — | — | — | — | ar bn zh +20 | Permissive | Not stated | mixed | 2026-01 | ||
CVSS-XGris et al. | — | ~16,070 hsynthetic target speech | — | — | — | — | en +28 | Non-commercial | Consented | mixed | 2026-09 | |
| — | 8.76 h | — | — | — | — | en | Non-commercial | Mixed | mixed | 2024-06 | ||
InteractSpeechInteractSpeech authors (EMNLP 2025) | — | 150 haudio not released | — | — | — | — | en | Custom | Consented | mixed | 2025-11 | |
| — | 4,064 hconversational configs | — | — | — | — | en fr de +12 | Custom | Mixed | mixed | 2026-07 | ||
| — | 1,145 h | — | — | — | — | en zh | Custom | Mixed | mixed | 2025-09 | ||
DuplexGenDuplexGen authors (Kim et al.) | — | 1,015 h | — | — | — | — | en | Permissive | Public media | synthetic | 2026-07 | |
VoxDialogueVoxDialogue authors (ICLR 2025) | — | — | — | — | — | — | en | Permissive | Public media | synthetic | 2025-05 | |
| — | 2,164 h | — | — | — | — | en | Permissive | Not stated | synthetic | 2025-04 | ||
Korean full-duplex synthetic (preview)Wi-Fi (Hugging Face) | — | preview only2,000.5 h claimed | — | — | — | — | ko | Permissive | Not stated | synthetic | 2026-08 | |
VoiceAssistant-400Kgpt-omni (Mini-Omni authors) | — | — | — | — | — | — | en | Permissive | Not stated | synthetic | 2024-09 | |
| Full-duplex | 5,446 h | — | avg 1.8 min | Customer service, Casual talk | — | hi en | Non-commercial | Consented | synthetic | 2026-08 | ||
| — | — | — | — | — | — | en | Non-commercial | Public media | synthetic | 2024-09 | ||
| Full-duplex | 57.6k h | — | — | — | — | en zh | Non-commercial | Not stated | synthetic | 2026-09 | ||
MoVE (paired EN/ZH emotional speech)Chen et al. | — | 1,056 hEnglish side | — | — | — | — | en zh | Non-commercial | Not stated | synthetic | 2026-04 | |
| — | 833 h | — | — | — | — | en zh ja +1 | Custom | Consented | synthetic | 2025-10 | ||
| — | 26.7k h | — | — | — | — | en | Custom | Public media | synthetic | 2026-08 | ||
DEMANDInria / University of Edinburgh | — | ~5 h | — | — | — | — | multi | Share-alike | — | — | 2013-06 |