AI Digital Human Lab

How do you play?

The game opens straight into a face-to-face conversation with Luna. Type in the chat bar or click the microphone and just talk: she answers in her own voice, with lip-sync, facial expressions and gestures. The icons on the right change her outfit, hair, face, voice, pose, motions, backdrop and language. Drag with the right mouse button to orbit the camera and use the mouse wheel to zoom. Optionally, click the robot button in the chat bar to connect your own AI agent.

How is AI used in the gameplay?

Generative AI is the core of the experience. Her replies are written live by a language model, spoken by a text-to-speech model in her own voice, and turned into lip-sync, facial expressions and gestures in real time. On request she writes and sings short original songs (lyrics by the language model, music by an on-device music model) and learns dances from MMD files (vocal separation and speech recognition models). The webcam mirror uses a pose-tracking model. The characters, outfits, animations and environments are hand-made 3D content, not AI-generated. Some background artwork in the store images was made with the help of AI image tools; the character shown in them is a real in-game render.

What technology powers the AI?

By default everything runs locally on your PC, with no account and no subscription: a Qwen3 language model through llama.cpp, Qwen3-TTS voices, ACE-Step for songs, Demucs and Whisper for learning songs, and MediaPipe for the webcam mirror. If you prefer, you can add your own OpenAI, Gemini or Grok API key, use your own ElevenLabs account for voices, or connect your own AI agent (for example Muse or Claude) through the built-in MCP server. The game is built with Unreal Engine 5 and MetaHuman.

How is content moderated?

Every line the player types and every line she is about to say passes through a deterministic safety filter that runs on your PC. It blocks sexual content involving minors, explicit sexual descriptions, hate slurs, and requests for instructions to hurt people or to make weapons or drugs. When it triggers, she declines with a calm fixed line, and the blocked text is never logged or uploaded. The built-in model is also instructed to stay friendly and respectful. Cloud providers you choose to connect apply their own moderation in addition.

Is there a way to report or block unwanted content?

Yes. You can interrupt her at any time, stop a reply, or clear the conversation, and cloud models or connected agents can be switched off in the settings. To report AI-generated content, use the Report button on this store page or contact us through allianceinterstellar.com. Please include a short description and, if possible, a screenshot. Every report helps us improve the filter.

What is included in Early Access?

The full core experience: real-time voice and text conversation with Luna and Orion, 74 facial expressions, on-device voices in 10 languages, a 31-language interface with optional dual subtitles, outfits and hairstyles, face sculpting with skin tone and makeup, poses and a motion library, learning dances from MMD files, original songs written for you, the webcam mirror, studio camera and lighting, and AI agent control through MCP (including photos and videos taken by your agent).

What is the current state of the product?

AI Digital Human Lab is fully playable. During Early Access we are expanding the wardrobe and the cast, making her voice and replies faster and more natural, and adding more ways for AI agents to direct the scene. Updates are delivered automatically through the Epic Games Launcher.

Will I have access to the product during the entire Early Access phase?

Yes. Your purchase includes the game during the entire Early Access phase and at full release, together with all updates to the base game.