Aether

Private AI chat,
on your own machine.

A clean chat interface for any Ollama model. Text, images and hands-free voice, and nothing leaves your device.

Get it on GitHub View the source

macOS, Windows or Linux · Python 3 and Ollama · MIT licence

The Aether welcome screen with a model picker and a message box

It can see

Attach, paste or drag in a picture and ask about it. Aether switches to a vision model on its own, then back again.

Aether describing an attached image

Just talk to it

Live mode is a hands-free conversation: it hears when you stop, answers out loud, then listens again. Talk over it to interrupt.

Live voice mode

A chat that feels finished

Replies stream in with a live speed meter, code blocks have one-click copy, and your conversations stay in a sidebar in your own browser.

A conversation in Aether with a code block

And the rest

  • Any model you haveQwen, Llama, Mistral, Gemma and more. Swap from a dropdown, no restart.
  • No dependenciesThe whole server is one short Python file using only the standard library.
  • No cloud, no keysNo sign-up and no API key. It talks to the Ollama on your machine and nothing else.
  • Reads replies aloudUsing your computer's built-in voices.
  • Speech stays localOptional voice input is transcribed on your device with Whisper.
  • Works on your phoneOpen it from a phone or tablet on the same network.

Install

  1. Install Ollama and pull a model: 
    ollama pull qwen2.5:7b
  2. Get Aether and start it: 
    git clone https://github.com/ZANYANBU/aether.git
cd aether
./run.sh
  3. Open http://localhost:8100 and pick your model.

Want it to understand images? Also run ollama pull qwen2.5vl:3b. Voice input is optional and needs faster-whisper; the README has the steps.