Open-source coding, built for Japan

Jouzu is our CLI agentic coding harness, with one-click access to the latest open-source coding models through the Shisa API. Served from Japan: your data stays private and is never used for training.

Learn More

Explore Our Solutions

Install

$ npm install -g jouzu

then run jz · needs Node.js 22.19+

Connect Jouzu Download for Windows

Jouzu Launcher v0.2.3 for Windows (x64) · signed by Shisa, Inc. Installs the desktop launcher together with the Jouzu runtime. Source code is public on GitHub.

Shisa Speech

Japanese voice AI that listens and speaks like a local.

Two halves of one conversation: speech recognition that understands your Japanese callers, and voices they trust when you answer — built together for contact centers, voice agents, and live transcription.

Speech → Text

Shisa ASR

The best and most affordable Japanese speech recognition

Industry-leading Japanese accuracy at a price that makes it practical to transcribe everything. Stays stable in noisy environments, with hotwords to steer transcription toward your own names and terminology. Stream in real time or process files in batch.

Top Japanese Accuracy
Low Cost at Scale
Real-Time & Batch
Hotwords for Your Terms

From $0.36 Price per hour

API Docs

Text → Speech

Shisa TTS

The best voices for your Japanese customers

The most local-sounding Japanese voices available — natural pitch and intonation, vetted by native speakers. Made for customer support, contact centers, and voice agents, where your customers should hear a voice that sounds like home. Low-latency streaming keeps every conversation natural.

Truly Local Japanese Voices
Vetted by Native Speakers
Built for Customer Support
Low-Latency Streaming

From $0.08 Price per 1K Characters — Flash Monolingual

API Docs

Shisa LLMs

Sovereign, powerful, small, and cheap

Japan's own bilingual models. Open weights let you run them on your own infrastructure and keep data in-country — or call them through our OpenAI-compatible API. Small enough to be fast and affordable, strong enough to lead Japanese benchmarks.

Sovereign & Open-Weight
•
Top Japanese Benchmarks
•
Small & Fast
•
OpenAI Compatible

From $0.20 Input per 1M tokens

API Docs

SHISA 2.1

Japan's Highest Performing LLM

Open LLM Leaderboard, Japanese subset

The AI Partner for Global Expansion

Built on best-in-class research technology and deep respect for Japanese culture

Accuracy

Best-in-class Japanese-English translation powered by proprietary LLMs

Our models are trained on extensive hand curated bilingual datasets, delivering nuanced translations that capture context, tone, and cultural subtleties with unmatched precision.

Customization

Tailored ASR, TTS, and fine-tuning for specific needs

Every business is unique. We support custom just-in-time context prompting and optional model fine-tuning to match your industry terminology, brand voice, and specific use cases.

Localization

Deep understanding of Japanese business and language nuances

More than translation - we understand Japanese business culture, formal registers, and regional variations to ensure your message resonates authentically.

Common Use Cases

See how businesses use our APIs

Customer Support

Translate support tickets in real-time, generate voice responses, and transcribe customer calls for multilingual customer service.

E-Learning Platforms

Create multilingual course content, generate audio lessons, and transcribe lectures for accessible education.

Content Localization

Translate websites, apps, and marketing materials with context-aware accuracy for the Japanese market.

Voice Assistants

Build bilingual voice interfaces with ASR for input, LLM for understanding, and TTS for responses.

Media & Broadcasting

Generate subtitles, dub content, and create voiceovers for international media distribution.

Gaming & Entertainment

Localize game content, generate character voices, and enable real-time player communication.

Pricing

USD ($)JPY (¥)

Latest models

View full model specs

Shisa 2.1 Flash

Fast and cost-efficient for everyday tasks

Input per 1M tokens$0.20

Output per 1M tokens$0.40

Shisa 2.1 Pro

Optimal balance of intelligence and speed

Input per 1M tokens$0.60

Output per 1M tokens$1.20

Qwen 3.7 Flash

New! Low-latency, multimodal model — hosted on Alibaba Cloud, Singapore

Input per 1M tokens$0.03

Output per 1M tokens$0.13

Qwen 3.7 Plus

New! Balanced multimodal model for coding and productivity — hosted on Alibaba Cloud, Singapore

Input per 1M tokens$0.40

Output per 1M tokens$1.60

Qwen 3.7 Max

New! Flagship model for complex agentic tasks — hosted on Alibaba Cloud, Singapore

Input per 1M tokens$1.65

Output per 1M tokens$4.96

GLM 5.2

New! Long-horizon reasoning and code generation — hosted on Alibaba Cloud, Singapore

Input per 1M tokens$1.40

Output per 1M tokens$4.40

API services

Translation

Best-in-class Japanese-English LLM translation

Price per 1K characters$0.02

Speech Recognition

Accurate speech-to-text optimized for Japanese

Price per hour$0.36

Text-to-Speech

Native sounding Japanese voice synthesis including support for multilingual and faster monolingual models

Price per 1K Characters — Flash Monolingual$0.08

Price per 1K Characters — Multilingual$0.10

Character count includes both input and output.

All prices are subject to applicable taxes.

Volume discounts available. Contact us for details.

Ready in under a minute

Install Jouzu, run jz, and start building.

Connect Jouzu

Start Building with Shisa APIs

Request your API key and make your first call in minutes. Full access to LLM, Translation, TTS, and Speech Recognition — no sales call required.

Get Started