About This Software

What is AIvy?

AIvy is a desktop mascot that brings a 3D character to your PC desktop—one you can actually talk to, powered by AI. Load your favorite VRM model and watch it come to life in the corner of your screen, chatting naturally like a companion who's really there. The name comes from "AI + ivy"—the gentle vine that quietly winds itself around your everyday life.

Build Your Own Companion, Your Way

The heart of AIvy is the freedom to combine the look, the mind, the voice, and the personality—piece by piece—to craft a companion that's truly yours.

Any VRM works: one you bought on a storefront, one you built yourself in VRoid Studio, or one from VRoid Hub. VRoid Hub integration shows each model's usage terms (avatar use, modification, redistribution, commercial use) for your consent before downloading. Files can also stay where they are and be referenced without copying them into the library.

Loading a model is only the start. Swap to alternate textures bundled with the model, toggle individual parts on and off, map the model's own expressions onto AIvy's emotions, set the idle expression, and tune lighting (ambient and directional intensity, direction, color)—all saved per model. There is also a beta feature for outfit swapping: load .unitypackage outfit assets, the format sold on storefronts, straight onto your model (outfits authored for the same base body).

Personality and speech style can be set from ready-made presets, or written freely in your own words to shape every nuance. You can also pick how it responds—empathetic, balanced, or candid—so your companion isn't just a yes-machine.

With support for 11 languages including Japanese, English, Chinese, and Korean, people around the world can spend time with their own one-of-a-kind companion in their own language.

The Mind Runs on Your PC

For the mind, AIvy is a local LLM runner built on llama.cpp. On first launch you pick from a dropdown an open-source language model that fits your language and download it once. From then on it runs fully offline, with no account, no API key, and no internet connection required. Your conversations are never sent anywhere.

You are not limited to the models listed in the app, either. Any GGUF file—say, one you grabbed from Hugging Face—works as AIvy's mind: pick it in settings or drop it into the models folder. There is no size cap, so a large model will run if your RAM allows.

In short, AIvy is the same kind of local LLM client as LM Studio, Jan, or GPT4All—just dressed up in a VRM character UI instead of a chat panel. The first-run download catalog is restricted to permissive licenses (Apache 2.0 / MIT) so the models are safe to use in commercial and streaming contexts.

The voice is delivered by your OS's built-in TTS. For something more characterful, there is Piper TTS: AIvy downloads and sets up both Piper itself and its voices from the settings panel, so there is nothing to install separately. Voices are available for every supported language, and you can preview one before downloading it. If you already have VOICEVOX, AivisSpeech, Sharevox, or VoiSona Talk, you can connect those to AIvy as well.

A Quiet Presence, Designed Not to Get in the Way

AIvy isn't built to be the assistant that keeps interrupting you—it's built to be the companion whose presence you simply feel. When you're focused, it watches over you in silence. In a quieter moment, it might softly check in with "Maybe time for a break?" or "Done with that video?"—a single line that fits the situation. And whenever you feel like talking, it's always there to answer. The mascot stays on top of every window but lets your clicks pass right through, so it never blocks your work. Every detail is tuned so it stays comfortable to have around, even on the longest workdays.

With these gentle gestures and the right kind of distance, AIvy adds a quiet color to your desktop life. Work, breaks, time alone—every part of your day gets a little warmer with AIvy by your side.

What Your Companion Can Do

It speaks up

• Proactive talk — you don't have to start the conversation. In a lull, or when you wrap up a task, it offers a line that fits the moment.
• Reads the room — calls, online meetings, games, videos, and streaming are detected automatically, and it stays quiet while they last.
• Active-app awareness — it recognizes the app you're using (game, browser, editor) and replies with topics that fit what you're doing now.
• A greeting at startup that accounts for how long you've been away.
• The occasional remark as sunrise or sunset approaches.

It helps out

• Alarms — at a time you set, the character tells you with its voice, not a sterile notification chime.
• CPU-load check-ins — when your PC stays under heavy load, it gently worries about you and your machine.
• Calendar — subscribe with an iCal (.ics) URL and it knows your schedule, reminds you of upcoming events, and can run through today's or tomorrow's plans. No account linking required; it simply reads the URL.
• Article recommendations — using the interests that come up in conversation, the software you've been using, and your region, it quietly suggests relevant web articles. After you read one it asks how it was, and learns your taste and preferred sources from your answer. Ask for "that article again" and it will remember and reopen it. Each of the 11 languages has its own local news sources. The feature is opt-in (off by default), and remembered words and software can be reviewed or deleted anytime.

It remembers and grows

• Long-term memory — it remembers your conversations, tastes, and habits, and weaves them into what it says next. It searches your past conversations for what relates to the current topic, so it gets to know you the more you use it.
• Backup and restore for settings and memory, so your companion survives a reinstall or a new PC.

You can talk to it

• Voice input — talk to it through your mic. Speech recognition also runs on your PC.
• Chat window — type to it while reviewing your conversation history.

No Servers. Always on Your PC

AIvy is a standalone app that runs on your PC alone. Everything—generating the conversation (local LLM), synthesizing the voice, recognizing your speech—happens on your machine, with no server of ours in between. Which means:

• Your conversations are never sent anywhere. Your privacy is protected.
• No account required.
• No subscription, no per-token billing, no running costs. Buy it once and it stays with you.

Note: article recommendations (fetching web articles), calendar integration (fetching your schedule), and downloading language and voice models do use an internet connection by nature. These are opt-in conveniences you can toggle on or off — the conversation, speech synthesis, and speech recognition themselves run offline.

Supported Engines

LLM (the mind)

• llama.cpp-based local inference — conversations stay on your PC
• On first launch, a dropdown suggests a model that fits your language
• Japanese / English / European languages / Russian: Gemma 4 E4B by default, with Sarashina 2.2 3B, Qwen 2.5 3B and 7B, Mistral Nemo 12B, and Phi-3.5 mini available
• Chinese / Korean: Qwen 2.5 3B by default, with Qwen 2.5 7B and Gemma 4 E4B available
• More models (Gemma 4 12B, Qwen 3 8B, Qwen 3.5 4B/9B, Phi-4 14B and others) can be downloaded from settings to match your PC and use case
• Every model in the catalog uses a permissive license (Apache 2.0 / MIT and similar). You can also bring your own .gguf — respecting its license is then up to you

TTS (the voice)

• OS Built-in TTS (Windows SAPI)
• Piper TTS — downloaded and set up by AIvy from the settings panel, both the engine and the voices. Nothing to install separately, and voices can be previewed before download
• Optional local TTS integrations (VOICEVOX / AivisSpeech / Sharevox / VoiSona Talk) — if you have installed them separately

Speech recognition

• Runs on your PC. The model is downloaded from settings only when you want to use voice input.

Important Notes

• Please use VRM models you have proper rights to

Streaming, Videos & Let's Plays

You are free to stream, record, and share content created with AIvy. When you do, please follow the license terms of the VRM model, the TTS engine (VOICEVOX, AivisSpeech, etc.), and the language model (Gemma, Qwen, Sarashina, etc.) you actually use. Whether commercial use, monetization, and on-screen credits are required depends on each VRM, engine, and model's own license — not on AIvy.

AI Generated Content Disclosure

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

AIvy is a next-generation desktop mascot whose character dialogue is generated by AI on the fly.

When you talk to AIvy, or when the character spontaneously speaks up about something happening on your PC, the words aren't pre-written scripts — the AI thinks them up and replies in real time. The generated dialogue is then read aloud by a text-to-speech engine, synchronized to the character's lip movements.

[What the AI generates]
- The character's dialogue (text)

[What the AI does NOT generate]
The 3D character itself (VRM model), facial expressions, animations, UI, and sound effects are all pre-authored assets. No images, video, or music are generated by AI during use.

[About the AI models]
Dialogue generation (default, local inference): AIvy is a local LLM runner built on llama.cpp (the same kind of architecture used by LM Studio, Jan, GPT4All). On first launch you pick from a dropdown the open-source language model best suited to your language — Sarashina 2.2 3B (MIT) for Japanese, Qwen 2.5 3B (Apache 2.0) for the others — and download it once from HuggingFace. From then on inference happens entirely on your PC, and your conversations are never transmitted externally. The in-app catalog only offers models under permissive licenses (Apache 2.0 / MIT).

Dialogue generation (optional, cloud LLM): For users without local-LLM-capable hardware, AIvy additionally offers optional cloud LLM integration with OpenAI, Anthropic Claude, and Google Gemini. Network traffic occurs only when the user has explicitly registered their own API key with the chosen provider in the in-app settings. API keys are stored with OS-level encryption (Windows DPAPI) on the user's device and are never sent to any AIvy server (we don't run one). Usage costs are billed directly by the chosen provider. When this feature is not used, no external traffic occurs and inference runs entirely locally.

[Streaming / video creation]
You may freely stream, record, and share content created with AIvy. Just follow the license terms of the VRM model, each TTS engine, and the language model you actually use — commercial use, monetization, and credit requirements vary per VRM/engine/model.

[Things to know]
AI responses vary every time, depending on the model you choose, the personality you configure, and what you say. For the rare case of an unexpected response, you can switch personality presets, change models, or pause speech at any time.