East Asia · Japan
Japanese conversational speech dataset
500 hours of spontaneous Japanese conversations recorded in Japan across 50 vetted native speakers. Dual-channel audio with time-aligned transcripts, speaker metadata (age, gender, region, device), and full chain-of-consent documentation. Licensed for commercial ASR, TTS, and voice AI training.
Dataset specifications
| Language | Japanese |
|---|---|
| Recorded in | Japan |
| Speech type | Spontaneous two-person conversations |
| Volume | 500 hours |
| Speakers | 50 vetted native speakers |
| Audio | Dual-channel (one channel per speaker) |
| Transcripts | Time-aligned, human-verified |
| Metadata | Age, gender, region, recording device |
| License | Commercial (training & evaluation) |
| Price | $95 per hour of audio |
Need more hours, another dialect or your own recording setup? We collect speech data to spec.
Built for
Training and evaluating speech models that must handle real Japanese speech:
- Automatic speech recognition (ASR / STT)
- Text-to-speech & voice synthesis prosody
- Voice agents and IVR for Japan
- Speaker diarization & turn-taking models
- Speech LLM / audio-understanding fine-tuning
Why conversational data?
Models trained on read speech consistently underperform on spontaneous audio. Disfluencies, overlap, and natural speech rates aren’t in the training distribution. This dataset closes that gap for Japanese. Read more in our guide to conversational vs. scripted speech data.
Free samples
Audit the Japanese data yourself
Share your company email and we’ll email you representative audio, matching transcripts, and the metadata schema. Check audio quality, transcript accuracy, and speaker diversity before you spend a dollar.
Public alternatives
Public Japanese speech data vs. this dataset
| Source | What it is | License | Commercial training |
|---|---|---|---|
| Corpus of Spontaneous Japanese (CSJ) (opens in a new tab)NINJAL (National Institute for Japanese Language and Linguistics) | 661 hours of Japanese speech, about 90% monologues, the rest conversation, recitations and re-readings. | NINJAL CSJ user agreement; separate academic, general and commercial fee tiers | Paid license |
| CALLHOME Japanese Speech (LDC96S37) (opens in a new tab)Linguistic Data Consortium (LDC) | About 49 hours from 120 unscripted telephone conversations between native Japanese speakers, placed from North America; transcripts are a separate LDC release. | LDC User Agreement for Non-Members (fee applies; non-commercial research and education only); commercial use requires LDC for-profit membership | Paid license |
| ReazonSpeech (opens in a new tab)Reazon Holdings (reazon-research) | Large corpus of natural Japanese speech collected from terrestrial television streams. | CDLA-Sharing-1.0, with use limited to information analysis under Japanese Copyright Act Article 30-4 | Unclear |
| Common Voice Scripted Speech (Japanese) (opens in a new tab)Mozilla Common Voice (Mozilla Data Collective) | Read speech recordings in Japanese from Mozilla's Common Voice project. | CC0 1.0 | Allowed |
CSJ is mostly monologue, ReazonSpeech is TV audio, Common Voice is read, and CALLHOME's LDC phone calls were placed from North America. This dataset is two-person conversation recorded in Japan.
Licenses as published by each source, checked October 2026. Check the current terms before use.
FAQ
Frequently asked questions
What does the Japanese speech dataset include?
500 hours of spontaneous two-person conversations from 50 vetted native speakers recorded in Japan, delivered as dual-channel audio with time-aligned transcripts, speaker metadata (age, gender, region, device), and consent documentation.
What makes Japanese hard for speech recognition?
Japanese is written with three scripts and no spaces, so transcripts need clear rules for kanji, kana and word boundaries before accuracy can be measured. Pitch accent, many homophones and the shift between polite and casual speech make conversation very different from read text.
Is there free Japanese speech data?
Yes, Common Voice is free for commercial use. The Corpus of Spontaneous Japanese and CALLHOME Japanese Speech need a paid license, and ReazonSpeech has unclear terms. This dataset is spontaneous two-person conversation recorded in Japan, with speaker metadata, one channel per speaker and signed commercial-use consent.
Can I use this Japanese dataset commercially?
Yes. The dataset is licensed for commercial use, including training ASR, TTS, and other voice-AI models. Every speaker has signed explicit commercial-use consent, and we provide the full consent chain for your legal review.
How much does the Japanese speech dataset cost?
Pricing is $95 per hour of audio. Volume discounts are available for full-dataset and multi-language purchases. Contact us for a quote.
Can I evaluate samples before buying?
Yes. Request samples with your company email, and we'll email you representative audio files and matching transcripts so your team can audit quality before committing.
$70/hr