Skip to content

ai-coustics/Project-Tyto-Real-Time-Demo

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

4 Commits
 
 
 
 

Repository files navigation

Tyto · Acoustics-aware voice agent (live demo)

A self-contained browser demo of Tyto, audio insight for voice AI. You talk to a live voice agent; in parallel Tyto scores your microphone in real time and the agent adapts to your acoustics: slowing down on bad audio, retuning turn-taking in a noisy room, and nudging you when something is fixable ("Could you turn the TV down?").

This is a single self-contained index.html. There is no backend and no build step: each visitor enters their own OpenAI and ai-coustics keys, which stay in their browser and go straight to those APIs over HTTPS.

Try it

Open the page, click Enter keys, paste your own keys, then click the mic and start talking.

Your keys are stored only in your browser (localStorage), used directly with OpenAI and ai-coustics, and never sent anywhere else. Clear them anytime from the Enter keys button.

What it does

Every couple of seconds Tyto scores your mic, and the agent reacts on three levels:

  1. Aware: a one-sentence "room note" is added to the agent's instructions, so it factors your acoustics into every reply.
  2. Tuned: in a noisy room it switches to a more patient turn-detection mode so it stops triggering on background sound.
  3. Reactive: when the risk is high and one cause dominates, the agent interrupts itself with a short spoken nudge. The Nudge sensitivity slider under the score bar tunes how eagerly that fires.

The agent also greets you with a fun prompt to get you talking, and you can ask "how do I sound?" any time.

Host it yourself

It is one file, so any static host works. For GitHub Pages:

  1. Put index.html at the root of a repo (or a docs/ folder).
  2. Settings → Pages → Deploy from a branch, pick the branch and folder, save.
  3. Share the https://<user>.github.io/<repo>/ URL.

The page must be served over HTTPS (GitHub Pages is), because the microphone and the in-browser WASM worker need a secure origin. Opening the file with file:// will not work.

Links

About

Tyto: Audio Insight for Voice AI

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages