A local-first, offline-capable AI Japanese language immersion engine. Step into interactive social roleplays with a private tutor and master natural usage across all JLPT levels. No cloud tracking, no subscriptions, just your own private language sandbox powered entirely by your hardware.

Sign in to add this item to your wishlist, follow it, or mark it as ignored

Coming Soon To Early Access

The developers of this game intend to release as a work in progress, developing with the feedback of players.

Note: Games in Early Access are not complete and may or may not change further. If you are not excited to play this game in its current state, then you should wait to see if the game progresses further in development. Learn more

What the developers have to say:

Why Early Access?

“Local-first AI language immersion is a rapidly evolving field. We are launching in Early Access to work directly with students and language enthusiasts to refine our "Vox Core" engine. Community feedback is essential to determining which social scenarios, linguistic guardrails, and educational features are most effective for real-world fluency.”

Approximately how long will this game be in Early Access?

“We plan to remain in Early Access for approximately 6 to 12 months. This window allows us to iterate on the AI’s conversational logic and expand the content library based on user learning patterns.”

How is the full version planned to differ from the Early Access version?

“The full version is planned to feature a significantly expanded library of social scenarios (targeting 100+ total), more specialized JLPT-specific learning paths, and enhanced neural speech synthesis choices. We also plan to implement deeper "Mastery" analytics alongside Steam-native features like Achievements and Cloud Saves.”

What is the current state of the Early Access version?

“The core "Vox Core" engine is 100% functional and offline-capable. The current version includes over 40 interactive social scenarios covering JLPT levels N5 through N1. Users can fully engage in conversational roleplay, utilize the integrated script-switcher, and track their vocabulary mastery through natural usage metrics.”

Will the game be priced differently during and after Early Access?

“We want to reward our early adopters. The price will be lower during Early Access, and we plan to gradually increase it as we ship major content updates and approach our full 1.0 feature set.”

How are you planning on involving the Community in your development process?

“We will be actively monitoring the Steam Community Hub to engage with our students and language enthusiasts. We plan to utilize the Steam Community forums to host regular polls where users can vote on the next social scenarios to be added (such as 'At the Gym' or 'Job Interview'). We also encourage users to report any unique linguistic edge cases they encounter; this feedback will be directly used to refine the AI’s tutoring logic and narrative stability, ensuring that our 'Vox Core' engine evolves to meet the needs of real learners.”
Read more
This game is not yet available on Steam

Planned Release Date: July 2026

Interested?
Add to your wishlist and get notified when it becomes available.
 
See all discussions

Report bugs and leave feedback for this game on the discussion boards

About This Game

The Anki Grind is Over. Welcome to Conversational Context.

Unlike standard language apps that function as digitized textbooks, Penko Vox: Japanese is a private, local-first immersion engine designed to bridge the gap between memorizing vocabulary and using it organically in real-world environments.

By merging a professional teaching methodology with a high-performance local LLM stack, Penko Vox acts as a private conversational "Sandbox", a safe space to make mistakes, test structural syntax, and build fluency without the social anxiety or recurring cost of a live human tutor.

The Vox Core: High-Fidelity Local Intelligence

Everything runs 100% offline on your own machine, driven by a proprietary 4-step local processing architecture:

  • Linguistic Perception: Our integrated speech-to-text layer captures your voice input instantly, while a deep morphological analyzer atomizes your sentences into particles, stems, and conjugations for real-time structural processing.

  • Cognitive Processing: An offline, highly optimized local LLM acts as your native tutor. Using adaptive narrative logic, the engine advances the scenario, injects world data, and prompts your response with high-level contextual awareness.

  • Educational Guardrails: The engine integrates native corrective formatting, modeling correct grammar and lexical patterns back to you inside the organic flow of dialogue rather than stopping to flag red execution errors or breaking the immersion loop.

  • Neural Synthesis: Responses are rendered with high-fidelity local neural synthesis, allowing you to hear the precise accentuation, native pacing, and pitch mapping of natural speech profiles.

Features for Power Users & Privacy Advocates

Total Data Sovereignty

Your voice recordings, database lookups, and conversational histories never touch an external cloud or telemetry server. Zero data leaves your machine, ensuring absolute privacy.

Ownership, Not a Service

Enjoy a self-contained immersion experience with zero server dependencies. Because the neural core resides entirely on your own hardware, you face no token limits, no artificial usage tiers, and zero monthly subscription costs. One purchase provides infinite practice, your tutor is always available, completely offline.

Interactive Stroke-Order Canvas

Master the fundamentals of the Japanese writing system with an integrated, high-fidelity practice environment. Use the interactive canvas to perfect your Hiragana and Katakana stroke order with real-time visual feedback and guided animations, ensuring your literacy keeps pace with your speaking fluency.

Mastery by Output Tracking

Ditch the static flashcards. Your progress is dynamically mapped via Natural Usage metrics. When you successfully use a vocabulary word or particle sequence correctly inside an organic conversation, your baseline Mastery percentage increases automatically.

Hardware & Architecture Notice

This software runs a state-of-the-art Generative LLM model completely on consumer hardware. Performance is fully dependent on your local CPU/GPU capability. AVX2 instruction support is strictly mandatory for the local inference binaries to execute.

AI Generated Content Disclosure

The developers describe how their game uses AI Generated Content like this:

Penko Vox: Japanese uses local generative AI to power its immersive language learning experience. An integrated Large Language Model (LLM) acts as a native Japanese tutor, generating real-time dialogue based on user input, while a Neural Text-to-Speech (TTS) engine synthesizes high-fidelity Japanese audio for listening practice.

System Requirements

Windows
SteamOS + Linux
    Minimum:
    • Requires a 64-bit processor and operating system
    • OS: Windows 10 (64-bit)
    • Processor: Quad-core CPU with AVX2 support (Intel Core i5-8400 / AMD Ryzen 3 3100)
    • Memory: 8 GB RAM
    • Graphics: Intel UHD Graphics 630 / AMD Radeon Vega 8
    • DirectX: Version 11
    • Storage: 15 GB available space
    • Sound Card: Windows Compatible Sound Card
    • Additional Notes: AVX2 support is mandatory for local AI inference. SSD is highly recommended for faster model loading.
    Recommended:
    • Requires a 64-bit processor and operating system
    • OS: Windows 11 (64-bit)
    • Processor: Hexa-core CPU or better (Intel Core i7-12700 / AMD Ryzen 5 5600G)
    • Memory: 16 GB RAM
    • Graphics: NVIDIA GeForce RTX 3060 / AMD Radeon RX 6600 (4GB VRAM recommended)
    • DirectX: Version 12
    • Storage: 20 GB available space
    • Sound Card: Windows Compatible Sound Card
    • Additional Notes: 16GB of RAM allows for the most fluid "Time to First Token" performance when using the 9B parameter model.
    Minimum:
    • OS: SteamOS 3.0 or Modern Linux (Fedora 40+, Ubuntu 22.04+)
    • Processor: Quad-core CPU with AVX2
    • Memory: 8 GB RAM
    • Graphics: Integrated or Dedicated GPU with Vulkan support
    • Storage: 15 GB available space
    • Sound Card: PulseAudio or PipeWire compatible
    • Additional Notes: On Linux, ensure your user has permissions to access hardware acceleration (render group).
    Recommended:
    • OS: SteamOS 3.5+ or Fedora 43
    • Processor: AMD Ryzen 5 5600G / Intel i7 equivalent
    • Memory: 16 GB RAM
    • Graphics: Dedicated GPU with 4GB+ VRAM (AMD/NVIDIA)
    • Storage: 20 GB available space
    • Sound Card: PulseAudio or PipeWire compatible
    • Additional Notes: Optimized for high-speed learning. With 16GB of RAM and an SSD, your AI tutor will respond almost instantly (under 3 seconds), providing a fluid, life-like conversational experience without any lag.
There are no reviews for this product

You can write your own review for this product to share your experience with the community. Use the area above the purchase buttons on this page to write your review.

Review Filters