Coming soon

About This Game

Note: This game is available in Japanese only (interface, voice and text). Conversations require speaking Japanese.

Desktop Friends (デスクトップフレンズ) is a communication game in which a childhood friend lives at the edge of your PC screen and talks with you by voice.

A friend you can simply talk to

Just speak into your microphone and the character answers in her own words. There are no fixed dialogue choices — talk about your day, your favorite things, or anything on your mind.

Two childhood friends

• Tsumugi — your junior by a year. Teases you politely, but is caring at heart. Loves stars and fortune-telling.
  
• Hiyori — your senior by a year. Laid-back, kind, and quick to praise you. Loves sweets and horror movies.

She remembers you

She remembers your favorite things, your plans and the name you like to be called, and asks how things went on another day. Every day's conversation is kept in her diary.

Closer friends, new outfits

The more you talk, the higher your affection level becomes, unlocking new outfits and gradually changing how she reacts.

Comments on your screen (optional, off by default)

With "Screen reactions" turned on, she sometimes glances at your screen and comments on the game or video you're watching. It is off by default; an explanation and consent screen appears before turning it on. Screenshots are never saved, and nothing is captured while password managers, banking or login screens are in front. You can turn it off any time from the menu or by saying 「見ないで」.

Timers and notes by voice

"Tell me in 10 minutes", "Wake me at 7:30", "Start a Pomodoro" — just say it. Notes, a shopping list and reminders work by voice too, and she reads them back when asked.

Your focus time, recorded

Pomodoro focus time is recorded for today, this week and in total, with your streak shown as a graph. While you focus, she quietly waits.

More stories the more you talk

28 short stories per character unlock as you talk and focus. Some ask you a question, and her reply changes with your answer.

Quizzes, fortune-telling and "Which do you like?"

150 quiz questions per character, star fortune-telling (Tsumugi) and sweets fortune-telling (Hiyori), playable by voice or buttons.

Seasons and anniversaries

Birthdays, the day you met, and seasonal events each have special lines.

Please note

• An internet connection is required, since conversations use an online AI service.
  
• A microphone is required for conversation.
  
• The game includes about 20 hours of conversation credit (varies with usage). Additional credit can be purchased in-game. Even with no credit left, timers, notes, quizzes, fortune-telling, stories, focus records and the characters' own remarks remain available (only speech recognition and AI conversation stop).
  
• Voices are generated locally on your PC. On PCs with a weak GPU, the text-only mode is recommended.

AI Generated Content Disclosure

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

[Pre-generated content]
- The character art of "Tsumugi" and "Hiyori" (standing pictures, outfits and facial expression parts) was created with image-generation AI, then selected, edited and split into parts by the developer.
- The characters' idle and conversation gestures were created from the character art with video-generation AI, then selected and edited by the developer.
- Some UI elements (such as the paper texture of the speech bubble) and some store images were created with image-generation AI and edited by the developer.
- The characters' prepared lines, short stories, quizzes, fortune-telling and "which do you like?" texts were drafted with text-generation AI, then reviewed, selected and edited by the developer.
- The characters' voices use AI speech-synthesis models ("Mao" and "Kohaku" for AivisSpeech, Aivis Common Model License 1.0).

[Content generated during gameplay]
- The characters' replies (an OpenAI large language model, used through the developer's server)
- The characters' voices (AI speech synthesis running locally on the user's PC)
- Transcription of the user's voice (speech-recognition AI)
- The characters' diary entries and summaries of what they remember from conversations
- Only if the user opts in: comments on screenshots of the user's screen (off by default)

To prevent inappropriate content, the character prompt instructs the AI not to engage with sensitive topics, every sentence is checked against a blocked-word list before it is spoken, and user messages are checked with the OpenAI Moderation API. Screenshots are never saved and are not taken while password managers, banking or login screens are in the foreground.