Alani · Voice
Say
“Hey Alani”
A voice assistant that listens, thinks and answers out loud. It runs on your own computer.
A conversation
What it looks like
Standing by...
Say “Hey Alani” to wake her.
An illustration of the on-screen widget: it shows whether Alani is standing by, listening, loading or speaking.
How it works
From wake word to answer
Wake
A small listener waits for “Hey Alani”. While idle it uses almost no power, and nothing heavy runs until you call her.
Listen
Speech recognition turns what you say into text. It ignores silence and background noise instead of guessing at them.
Think
A language model decides what you want and which tool to use, such as setting a reminder or checking the weather.
Act
She does it: adds the event, looks something up, rolls the dice. If she is missing a detail she asks for it, and if she cannot do something she says so instead of guessing.
Speak
A natural-sounding voice answers, faster than real time. When you are finished, she goes back to standing by.
Abilities
What she can do
Reminders
Set, list and delete reminders. Say the time loosely: “seven pm tomorrow” is enough.
Calendar events
Add, list and delete events with a title, a start and an end.
Answers
The current time, the weather, and quick searches of the web.
Games
Roll dice and draw cards.
Knowing when you are done
Say “that is all” or just wrap up naturally and she ends the conversation.
Manual entry
A “+” button opens simple forms to add a reminder or an event by hand, with no talking needed.
Design
Private and personal
On your computer
Speech recognition, the language model and the voice all run on the computer in front of you, not in someone else's cloud.
Your choice of voice
Four distinct voices, each created from a short sample recording.
Clear about its state
The widget always shows what she is doing: standing by, listening, loading, speaking or searching.
Works with Discord
Reminders and events you speak show up where you already look: in Discord.