Dialects, call audio, and natural conversation.
Japanese speech varies sharply by region, age and setting, and mora-timed prosody and pitch accent are hard to learn from scraped audio.
Describe the data you need. We will respond with feasibility, rights approach and next steps.
Speech recorded from native speakers across prefectures, with dialect and demographic metadata captured under consent.
Contact-center style call audio with consented participants or role-played scenarios, including keigo-heavy service language.
Professionally recorded, consent-licensed voice talent in specified styles, for text-to-speech and voice synthesis.
Acted and elicited emotional speech with nuanced Japanese affect categories, labeled by multiple native raters.
Age-diverse speech where mainstream ASR underperforms, collected under strict guardian and participant consent.
Speech captured in stations, cars, kitchens and shops for robust far-field recognition.
Environmental, urban and traditional-instrument audio for audio-generation and sound-event models, subject to per-source rights.
Not here? Describe the data you need. We source beyond this list.