桌面日常 Playtest · published · build 26018655
Over the past few weeks, we've been working on three major improvements:
Home Life: A new feature where you can watch two characters go about their daily lives.
Multilingual voices and faster responses: Added Chinese and English speech support while reducing the time it takes for characters to start speaking.
New animations and expressions: Completely reworked how characters move and express themselves during conversations.
We want our characters to do more than just answer questions on your desktop. We want them to move naturally when they speak, and to have lives of their own even when you're not chatting with them.
Click the house icon in the character status bar to enter a home shared by Koharune Ami and Tsukuyomi-chan.
They each have their own bedroom and share a living room and kitchen. They'll move around on their own, read, rest, cook, eat, and return to bed to sleep.
This feature is still in an early stage. Based on your feedback, we'll decide whether to continue developing it and which directions to explore next.
Click a character to focus the camera on her.
Physical interactions: Pat her head, gently pull her cheeks, tap her nose, scratch her chin, touch her ears, tap her shoulder, give her a high five, or shake hands.
Everyday interactions: Call her over, say goodnight, wake her up, or feed her while she's eating.
Context-sensitive reactions: Some interactions produce different responses depending on the character's current state, the time, and how many times you've touched her in succession.
Natural activity transitions: If she's reading, cooking, or eating, she'll put down what she's holding or finish her current action before responding to you.
Left mouse button + drag: Pan the camera.
Right mouse button + drag: Rotate the view.
Mouse wheel: Zoom in or out.
Automatic viewing: Let the camera move automatically.
Hide interface: Watch the characters' daily lives without UI distractions.
Indoor lighting, lamps, and wall clocks change according to your computer's local time.
Background music switches between morning, daytime, and nighttime tracks.
The music panel lets you choose tracks, enable playlist rotation or single-track repeat, and adjust the volume.
Your music preferences are saved automatically.
You can also set the home as a Windows animated wallpaper.
The desktop characters, Home Life, and animated wallpaper can each be assigned to different monitors.
Animated wallpaper support and multi-monitor compatibility are still being tested, so please let us know if you encounter any issues.
Koharune Ami and Tsukuyomi-chan can now speak Chinese, Japanese, and English, including sentences that mix all three languages.
Select the spoken language at the top of Settings → Character Voice.
Japanese remains the default language.
Once selected, characters will respond and speak in that language during regular conversations, Visual Companion interactions, and daily-life activities.
Interface and subtitle languages can be configured separately.
Mixed-language pronunciation: When Japanese names appear in Chinese sentences, the names will be pronounced in Japanese without incorrectly switching the surrounding Chinese text to Japanese pronunciation.
Character name pronunciation: Reduced unnatural pauses in the middle of character names.
Native pronunciation references: Added native-language pronunciation references for all three languages to reduce unnaturally prolonged speech, lengthy pauses, and skipped words.
Note: Nanairo Kokoro's voice is still generated using COEIROINK and currently supports Japanese only.
Regular conversations and daily-life activities on the desktop now use streaming speech generation and playback.
Characters no longer need to wait for an entire sentence to finish synthesizing before they start speaking.
The delay after sending a message has been significantly reduced.
Facial expressions now change in sync with the words being spoken.
Preloaded voice preparation: The game prepares voice synthesis in advance based on your selected spoken language, reducing the delay before your first conversation.
Automatic sleep mode: After a period of inactivity, the voice service enters sleep mode and releases most of its GPU memory.
Faster resumption: When a character speaks again, the service can resume without going through a full restart.
Significantly improved inference speed: Voice generation is now much faster on AMD and Intel GPUs, with performance approaching that of NVIDIA GPUs.
Fixed voice timbre issues: Resolved an issue where character voices sounded unnaturally deep when using AMD or Intel GPUs in DirectML mode.
Automatic device switching: When you change the speech synthesis device, the game automatically switches after the current sentence finishes playing. Restarting the game is no longer necessary.
Microsoft C++ runtime included: The game now includes the runtime libraries required for speech synthesis, fixing an issue that prevented the voice service from starting on some computers.
Fixed short-sentence playback errors: Resolved an issue where speech playback could fail for short sentences containing a space after an exclamation mark or question mark, such as "Hi! How are you?" or "你好! 今天怎么样?".
Previously, characters selected a predefined animation for each line of dialogue.
Now, the dialogue model choreographs character movements and expressions based on the dialogue itself.
The new system includes:
Poses and hand gestures.
Nodding and head tilts.
Eye movements and gaze direction.
Forward and backward body movements.
Facial expressions.
Smooth transitions between animations.
These movements are aligned with the corresponding spoken words, making the characters' performances feel more natural and connected to what they're saying.
Koharune Ami and Tsukuyomi-chan now use animations recreated specifically for their body proportions.
We've added more than 40 gestures, including:
Waving, making a heart with their hands, and pinky promises.
Handshakes, thumbs-up, and peace signs.
Covering their mouths while giggling and acting cute.
Apologizing with their hands together.
Flexing their muscles and counting on their fingers.
Saying goodnight with their hands against their cheeks.
Their standing idle poses have also been replaced, and they'll occasionally perform small idle movements.
Visual Companion responses and spontaneous dialogue now include animations and expressions.
Regular conversations through the official trial service also use the new animation system.
Visual Companion through the trial service currently supports facial expressions only.
Nanairo Kokoro also uses the new animation system, with expressions and movements synchronized to the words she speaks.
Adjusted her lip-sync animations to prevent her mouth from opening too widely.
Improved the shading on her clothing and her overall rendering quality.
Custom characters imported from VRoid Studio now automatically use the same animations, expressions, and rendering settings as Nanairo Kokoro.
These animations were designed for chibi-style characters, so certain hand gestures may not align perfectly on models with taller, more realistic body proportions.
Fixed an issue where the character status bar was not displayed for custom characters.
Two-character desktop mode temporarily disabled: The new animation system has not yet been adapted for conversations involving both characters. We'll consider bringing this feature back once compatibility work is complete.
Visual Companion setup improvement: When you open Visual Companion for the first time, the model configured during the setup guide will now be saved to Model Management and used directly as the Visual Companion model.
Moving forward, we'll continue expanding the characters' animation library, refining Home Life and voice synthesis, and listening to feedback from everyone using the game.
We hope that every word, expression, and little movement will make these characters feel a bit more like they're truly living on your desktop.