PrivacyWhat leaves the houseSheet 05
What stays in the house.
Vulpine runs on a computer in your house. This page says what stays there and what does not.
- This site
- no analytics, no cookies
- Your speech
- recognized at home
- The voice
- made at home
- Your agent sees
- text, never audio
- Test recordings
- never uploaded
- Mood reading
- off by default
-
01
This site
This site has no analytics, sets no cookies, and loads no third-party fonts or scripts. Everything a page loads comes from this site. The faces on the Fun! page are drawn in your own browser; that page opens no microphone and sends nothing back.
-
02
What stays at home
Speech is recognized on your own machine, and the voice is synthesized there too. Household audio never leaves your own machines: a check on every change to the source refuses a command that would copy audio anywhere else.
-
03
What your agent sees
Your agent gets the text of a turn that needs one, never the audio. The model it calls is usually a hosted one; the recognizer and the voice stay on your machine.
-
04
Recordings made to test it
To test a change, a developer sometimes reads a set of scripted turns aloud. Those recordings stay on that developer's own machines. They are never uploaded and never committed to the source; the measurements published from them carry counts only.
-
05
Who is speaking
Vulpine can compare a voice with the voice you enrolled and write the result down as evidence, but the gate that would act on it is off by default, so nobody is refused because of it.
-
06
Your mood
Reading your mood is opt-in only, and every part of it is off by default. When it is on, signs that a turn went wrong (a repeated request, a "no, I said", a failed tool) can make the next reply shorter or more careful. Vulpine never says how you sound, and it never names a mood back to you. No mood word, score or recording leaves the home machine; at most your agent gets a plain instruction, such as "keep replies short". Turning the switches off restores the behavior from before.