Speech Dataset
Our Advantages
REAL-WORLD SPEECH
Speech collected in authentic environments for practical model training.
AI-READY DESIGN
Dataset specifications aligned with ASR, conversational AI and voice model requirements.
GLOBAL COVERAGE
Multiple languages, accents, devices and recording environments.
MULTI-STAGE QA
Three-round quality inspection with project-specific acceptance criteria.
CONSENT & COMPLIANCE
Collection workflows designed around consent, privacy and applicable local requirements.
DELIVERY SUPPORT
Project support for dataset evaluation, delivery and technical questions.
Speech Recognition Dataset
English-China Children Speech Dataset
1,000 Speaker Number
217 Hours
English-China
English Speech Data-Conversation-India
Speech Hours : 75
Speakers : 75
Speech Style : Engineering Conversation
English Speech Data-India
Speech Hours : 415
Speakers : 750
Speech Style : Reading
English Speech Data -Sigpore
Speech Hours: 1100
Speakers : 1073
Speech Style : Reading
English Speech Data-China-APP
Speech Hours : 849
Device£ºLive data
Speech Style : Natural Language
English-US Call Center Speech Dataset-2
Speech Hours£º1020
Device£ºLive data
Speech Style : Natural Language
English Speech Data-Tanzania
Speech Hours : 227
Speakers : 404
Speech Style : Reading
English Speech Data–Kenya
Speech Hours : 235
Speech Style : Reading
Speakers: 462
English Speech Data-Conversation-Sigpore
Speech Hours£º163
?Speech Style £ºEngineering Conversation
162 Speakers
English China Children Speech Data
Speech Hours£º217
?Speech Style £ºReading
1000 Speakers
English Australia Speech Data
Speech Hours£º308
?Speech Style £ºReading
646 Speakers
Chinese Mandarin English co-switch2 Speech Data
Speech Hours£º2648
?Speech Style £ºReading
1468 Speakers
English-US Speech Dataset
865 Hours
1935 Speakers
Reading
English-US Call Center Speech Dataset
287 Hours
Age: >16 years old
Scene: Live
Chinese-Mandarin-English Speech Dataset Co-Switch
4089 Hours
8477 Participants
Reading
Chinese Mandarin Multimodel Generic Speech Dataset
Speech Hours : Each speaker shot 100 scripts£¬about 6-10minutes
Speakers : 500
Speech Style : Multimodel Spech Data
Talk To Us Now