v0.2.5
Added direct model download from Priva Studio's Hugging Face page
Loading game…
Developer posts and news about this game.
Added direct model download from Priva Studio's Hugging Face page
Fixed memory system not triggering correctly Increased memory extraction frequency
macOS is now supported Added collapse button on the right side, visible on hover Improved agent tool calling capability
Improved agent tool calling capability Fixed file editor save overwriting issue for skill
Added agent skill: Madato (read XLSX) Improved Markdown UI rendering Improved built-in agent skill: describe PDF
Added built-in agent skill: describe image Added built-in agent skill: describe PDF Added image and PDF file path support in history Updated Native Vision from Qwen2-VL to Qwen3-VL
Added support for Qwen3.5 and Qwen3.6 models Qwen3.5 is now the default model (Qwen3 removed) Added image input support for visual analysis Added Native Vision toggle (enable/disable) for clearer control GPU layer offloading (auto splits model layers between VRAM and RAM)
Added file editor for system prompt, memory, history, and skill Added interface language support: Simplified Chinese, Traditional Chinese, Japanese Added built-in skill: Everything
Support jepg texture
Look-At Mode: You can now choose what the character tracks with her gaze — either the cursor or the camera — switchable from the settings menu Body Tracking: The character's body now turns naturally along with her gaze direction, making interactions feel more alive and dynamic, i
Added Steam Workshop support. Players can now upload, share, and browse community-created VRM models, system prompts, and voices.
Added Agent Skill System — place a SKILL.md and executable in assets/skills/ to add custom skills Added Body Part Detector — hover over items in the Body Part Detector list to highlight the corresponding part on the VRM model