A private, local AI companion for Windows

1. What runs on your PC

Local chat generates replies using a model installed on your computer. Local speech recognition turns microphone input into text, local voice generates spoken replies, and local screen vision processes the screen content you choose to share. Your companion’s memories and settings are stored on your drive.

These features do not need a cloud subscription. You can keep chat, voice and vision local, or choose cloud processing separately for supported features. A 3D avatar on the desktop does not by itself tell you where an AI model runs; your model and feature settings do.

2. What works offline after setup

The website version saves an offline entitlement after successful activation. Ordinary local use on that activated installation does not require a recurring cloud subscription. Keep the installed models and activation data when preparing to use it without a connection. This guide describes the direct Windows download; store activation follows the store’s own rules.

3. What stays local, and what can leave

With local processing selected, your chat, locally generated speech, memories and screen observations are not sent to Asiden Labs for AI processing. We do not receive a backup of your local conversation history.

That does not mean an online installation makes no network requests: model downloads, update checks, licensing, purchases and deliberately enabled online features use the internet. Licensing traffic is separate from conversation content.

Local storage also means you are responsible for access to your Windows account and for any backups you make. Other software, a shared Windows session or a backup service you use may have access to files on that PC. See local processing in the Privacy Notice.

4. What changes when you choose cloud AI

Managed cloud chat currently offers models through Venice AI and OpenRouter. The available names and Spark costs appear in the app and can change. Cloud speech uses ElevenLabs; cloud vision uses the selected supported model route. See the local and cloud plan comparison before choosing.

• Cloud chat: your message and the context needed to answer can leave the PC. That can include recent conversation, relevant memories, companion instructions and enabled screen context.
• Cloud voice: reply text is sent for speech generation.
• Cloud vision: the selected screenshot and accompanying text are sent for image analysis.

Managed requests pass through Asiden’s proxy and the selected provider chain. Connections are encrypted in transit; this is not end-to-end encryption. Providers process data under the route’s applicable terms and retention settings. Keeping memories on disk does not prevent a relevant memory being included in a cloud prompt.

The app identifies cloud processing and requests consent before managed cloud use. Return each feature to local to stop sending future AI requests for that feature. Turning cloud chat off does not turn separately enabled cloud voice or vision off. The cloud-processing section gives the full details.

5. Your memories and their limits

Asiden saves useful facts from conversations so later replies can refer to things you have shared. The chosen personality and learned memories are separate: learning about you is not intended to rewrite the character you chose.

Memory can be incomplete or wrong. You can use the app’s memory controls, including a deliberate reset, and should keep your own backup if you need to preserve local data. A reset removes that companion’s learned conversation history; it does not remove copies you made elsewhere or reverse past cloud requests. Read how memory works in everyday use.

6. Cost, setup and hardware

The website Unlock is a one-off $10 during early access. Local conversation has no message allowance and no monthly fee. Buying it still involves checkout and an emailed activation key; “local” does not mean the app is free to download and use without an Unlock.

Plan for Windows 11 64-bit, 16 GB RAM, an NVIDIA GPU with at least 8 GB VRAM and 25 GB free storage under the published minimum. Model choice changes memory use and speed. A game, screen vision and a local language model can compete for the same GPU, so replies may slow down under load. Check the full hardware requirements before buying and follow the setup walkthrough.

7. When this may not suit you

• Asiden is for Windows 11 PCs; there is no phone, macOS or Linux version.
• A machine below the published requirements may not provide a usable local experience. Cloud processing changes where work is done, but does not remove every local hardware requirement.
• Asiden is early access. Response quality, memory recall and screen understanding can make mistakes.
• It is an AI companion for adults, not a therapist or crisis service. See safety and support boundaries.

8. Common questions

Can I use an offline AI companion without a subscription?

Yes. After the one-time website Unlock, activation and local setup, Asiden’s local features do not need a monthly subscription. Cloud features are optional and need internet access.

Do you train on my local conversations?

Asiden Labs does not receive local conversations to train on. Cloud processing has separate provider terms, described in the Privacy Notice.

Can I move my existing companion here?

There is no conversation-history import from another companion service. A new Asiden companion starts learning from your conversations with it.