Draggy
What it can do
Voice mode
Talk to it, and have it answer out loud.
Talk in the Chat sidebar is a spoken conversation, not dictation. It listens continuously, works out when you have finished, answers out loud, and lets you cut in.
It is beta, and that is honest. It wants a decent GPU and a headset. Over laptop speakers it will occasionally answer itself.
The orb
| What it does | What it means |
|---|---|
| Grows with your voice | Listening |
| A dot with a rotating arc | Thinking, or searching |
| Pulsing steadily | Speaking |
| An outline | Microphone muted |
Keys
| Key | Action |
|---|---|
Esc | End the conversation |
M | Mute the microphone |
C | Show the transcript |
Space | Skip the rest of the reply |
Voices
System uses the voices your operating system has: instant, and they sound like your operating system. Natural is a neural voice, about 90 MB on first use, English only, and considerably better. Speed runs from 0.9x to 1.25x.
Everything spoken stays local. Recognition runs with Whisper on your machine, and audio is never uploaded or written to disk. Web searches are the only thing that leaves.
When it goes wrong
- It interrupts itself. Its own voice is reaching the microphone. Use headphones.
- It is slow. Watch the millisecond counter. Over 1500 ms means the talk model is too big.
- "Basic detection". The neural voice detector did not load; background noise fools the simpler one.