- 注文状況を確認したいです。
Horizon · Radiance
Build AI voice agents that can listen to Japanese, understand customer requests in real time, and respond through natural Japanese speech across web, mobile, and telephony experiences.
Use Shunya Horizon and Radiance as the speech layer for Japanese customer support, order tracking, retail, travel, notifications, enterprise applications, and conversational AI - with streaming transcription, native Japanese text input, expressive TTS, telephony audio, speaker-aware transcripts, and post-call intelligence.
注文状況を確認したいです。
承知しました。注文番号を教えていただけますか?
Japanese speech · Horizon and RadianceJapanese AI Voice Agent
A Japanese-speaking user may ask:
The application needs to:
Shunya supports this voice-agent architecture for Japanese speech workflows.
| Capability | Japanese voice agent |
|---|---|
| Japanese Speech Recognition | |
| Real-time Streaming STT | |
| Japanese Text to Speech | |
| Native Japanese Text Input | |
| Kanji / Kana Input | |
| Streaming TTS | |
| Batch TTS | |
| Expression Control | |
| Speed Control | |
| Telephony Audio | |
| Speaker Diarization | |
| Speaker Identification | |
| Word-level Timestamps | |
| Intent Detection | |
| Sentiment Analysis | |
| Summarisation | |
| API Integration |
Try a Japanese voice conversation
Example
注文状況を確認したいです。
承知しました。注文番号を教えていただけますか?
48217です。
Another customer may ask:
Japanese speech to text
Japanese speech can appear in:
Japanese is included in Shunya's broader global speech-recognition coverage.
The transcript becomes the application input.
Real-time Japanese STT
Voice agents need transcription during the conversation. Shunya's speechstack supports WebSocket streaming for live voice recognition.
Use real-time STT for
The application does not need to wait until the call ends before using the transcript.
Japanese writing system
Normal Japanese text combines:
Kanji
Used extensively for names, concepts, nouns, and verb roots.
Hiragana
Used for grammatical forms, particles, and native words.
Katakana
Frequently used for foreign words, brands, technology, and imported terminology.
A Japanese AI voice application needs to work with real Japanese text rather than treating the language as one simple script. Shunya supports native Japanese text input including Kanji and Kana for Radiance.
Japanese + English terms
Japanese customer experiences frequently include:
Example
Shunya API を使ってリアルタイム音声を生成できます。
Another
Premium プランへの upgrade が完了しました。
Shunya supports Japanese text containing English and Latin-script terms.
Production applications should test the actual brand names and foreign terminology they expect users to hear.
Radiance for Japanese
Radiance includes Japanese within Shunya's extended world-language speech-synthesis coverage.
This generated speech can be used in:
Streaming Japanese TTS
A prerecorded Japanese announcement can wait. A conversational AI agent cannot.Radiance supports Japanese streaming synthesis over WebSocket.
<350 msFirst-audio latency supported for Japanese streaming TTS.
Streaming is suited to
Japanese TTS voice availability
Radiance Indic publishes 46 named voices across Indic languages and English. Japanese uses the world-language synthesis path instead.
World languages such as Japanese use the built-in speaker exposed by your Radiance deployment rather than a published dedicated Japanese male/female pair in the Indic voice catalogue.
Capabilities to validate
unless your deployment exposes them.
Supported position
Control Japanese speech style
Radiance documents 11 expression styles. Examples include:
Customer support
Announcement
Media / learning
The application can select an appropriate style according to the experience.
Control Japanese speaking speed
Radiance supports speed control from 0.25× to 4.0×.
Instructions, accessibility, and learning content.
Everyday conversation.
Concise information workflows.
The application can configure speech speed without changing the underlying response text.
Japanese audio formats
Radiance documents seven output formats: PCM, WAV, MP3, OGG Opus, FLAC, μ-law, and A-law.
Real-time pipelines
Web
High-quality audio
Telephony
One Japanese TTS platform can serve multiple delivery channels.
Japanese telephony
A Japanese-speaking customer may interact through:
Generated Japanese TTS supports μ-law and A-law formats for telephone-oriented output.
Customer support
A customer might ask:
Shunya handles speech. The support system handles the customer issue.
Order tracking
The commerce system remains the source of truth for the actual shipment status.
Appointment workflows
The scheduling system remains responsible for actual availability and appointment changes.
Retail
Retail is a supported Japanese TTS use case.
Possible experiences include:
Travel and hospitality
Travel and hospitality are supported Japanese TTS use cases.
Voice applications may support:
The connected travel or hospitality application remains responsible for business data and actions.
Notifications
Radiance can generate Japanese audio for:
Notifications are a supported Japanese TTS use case.
The notification can be pre-generated or produced dynamically depending on the workflow.
Enterprise applications
The same enterprise application may support:
Japanese becomes another voice interface around the existing software rather than a separate product architecture.
Human handoff
The application may transfer the interaction when:
The surrounding application controls escalation.
Speaker diarization
If a customer and human support agent speak in the same call, speaker diarizationcan separate the participants.
With diarization
SPEAKER_00 注文状況を確認したいです。
SPEAKER_01 注文番号を教えていただけますか?
SPEAKER_00 48217です。
Speaker-aware transcripts support:
Speaker identification
Speaker diarization separates voices. Speaker identification can matchregistered speaker profiles where configured.
Before
確認します。
After
確認します。
Unknown customers can remain anonymous labels.
Word-level timestamps
Verbose transcription can include, for individual words:
Use for
The timestamp layer makes a completed Japanese transcript more useful than plain text alone.
Post-call intelligence
Shunya's broader STT workflow documents batch intelligence fields including:
These enrichments should be positioned as post-call/batch capabilities rather than automatically live during every agent turn.
Intent
An organisation may define intents such as:
The organisation defines the taxonomy.
Sentiment and emotion
A customer may say:
何度も問い合わせています。
The transcript contains the actual statement. Batch intelligence can add:
Sentiment
Emotion
Dominant emotion associated with transcript segments.
Potential uses include:
These signals should complement the transcript rather than replace explicit customer meaning.
Call summaries
A completed call may contain:
Shunya's broader batch intelligence layer documents summarisation.
Japanese voice cloning and cross-lingual experience
Shunya's public Radiance experience demonstrates a cross-lingual voice flow including Japanese.
ImportantFor production voice-cloning availability and configuration, use the capabilities enabled for the specific Shunya account or deployment. Do not assume every deployment exposes the same cloned-voice configuration.
Japanese is not translation
Speech recognition, speech synthesis, and translation are separate capabilities in the Shunya product family.
Supported.
Supported.
Enable when your deployment exposes the Japanese translation pairs you need.
Cross-language translation through Parallax is separate from Japanese STT and TTS:
Japanese voice AI does not require translation when the interaction remains Japanese end to end.
Japanese + global voice AI
Shunya's Japanese TTS material places Japanese alongside supported world languages such as:
The language changes. The underlying business application does not need to be completely rebuilt.
Developers
Batch
Streaming
The existing Japanese TTS developer material documents batch HTTP synthesis, streaming WebSocket synthesis, and language code ja.
TTS configuration. Choose:
Important voice noteDo not hard-code an invented Japanese voice name. Use the voices available in the active Shunya deployment.
Real-world Japanese voice agent use cases
Provide spoken Japanese customer-service interactions.
Understand shipment questions and return order information through connected systems.
Build Japanese-speaking kiosks and product assistants.
Provide Japanese voice assistance and information.
Handle booking and scheduling conversations through connected applications.
Generate spoken reminders, updates, and alerts.
Add Japanese voice to existing business systems.
Create Japanese conversational assistants.
Provide spoken alternatives to written Japanese interfaces.
Generate dynamic Japanese dialogue where appropriate.
These use cases are consistent with the Japanese TTS applications Shunya supports for production voice AI.
Built for production Japanese voice AI
A production Japanese AI Voice Agent may need:
| Requirement | Shunya Japanese stack |
|---|---|
| Japanese STT | |
| Streaming STT | |
| Japanese TTS | |
| Native Japanese Input | |
| Kanji / Kana | |
| Streaming TTS | |
| Batch TTS | |
| First Audio <350 ms target | |
| Expression Styles | 11 |
| Speed Control | 0.25×–4.0× |
| Audio Formats | 7 |
| Telephony Output | |
| Speaker Diarization | |
| Speaker Identification | |
| Word Timestamps | |
| Intent | |
| Sentiment | |
| Summarisation | |
| API Integration |
One end-to-end Japanese conversational experience.
Questions
A Japanese AI Voice Agent is a conversational application that understands spoken Japanese, processes the request through an AI or business application, and returns spoken Japanese.
Yes. Japanese is included in Shunya's global speech-recognition coverage.
Yes. Shunya's broader Horizon platform supports WebSocket streaming for live voice-agent workflows.
Yes. Japanese is included in Radiance world-language synthesis coverage.
Use the standard Japanese language code ja in Radiance requests.
Japanese TTS supports native Japanese text, including normal Japanese writing using Kanji and Kana.
Radiance Indic publishes named voices for Indic languages and English. Japanese uses the world-language synthesis path rather than a dedicated Japanese male/female voice pair in that catalogue.
Yes. Streaming Japanese TTS is supported over WebSocket.
Japanese streaming TTS documents first audio in under 350 ms.
Yes. Radiance supports 0.25× to 4.0× speed control.
Yes. Radiance documents 11 expression styles.
PCM, WAV, MP3, OGG Opus, FLAC, μ-law, and A-law.
Yes. μ-law and A-law output formats are supported for telephony environments.
Yes. Japanese application text may include Latin characters and English terminology. Test pronunciation with the text your users actually speak and type.
Yes. Shunya's broader automated speech recognition platform supports speaker diarization.
Registered speaker profiles can be combined with speaker identification where configured.
Yes. The broader Horizon structured-output workflow supports word-level timing.
Yes. Summarisation is part of Shunya's supported batch intelligence layer.
Yes. Intent, sentiment, emotion, and related enrichment are available in supported batch STT workflows.
Yes. Japanese customer support is a supported Radiance use case.
Yes. Retail is a supported Japanese TTS use case.
Yes. Travel and hospitality are supported use cases.
Yes. Order-status conversations appear in the Japanese voice-agent examples.
No. Parallax's supported language set does not list Japanese, so English↔Japanese translation through Parallax should not be assumed.
Japanese automated speech recognition and speech synthesis are supported end to end in Japanese. Add Parallax translation only when your deployment exposes the language pair you need.
Yes. Japanese fits into Shunya's broader global speech platform alongside other languages.
Get started
Let Japanese users speak naturally. Use Horizon to turn spoken Japanese into machine-readable text. Use your AI or business applicationto understand and act. Use Radiance to respond through natural Japanese speech. Test customer support, order tracking, retail,travel, notifications, Japanese-English terminology, telephone audio, and the real conversations your users actually have.