Beyond Voice Chat

A voice front end for an AI model you run yourself. Speak, and the answer comes back as a voice, with a running transcript of both sides.

Web browser
Your own GPU computer
React
TypeScript
Web Audio
Qwen3-Omni

Not released yet In development

Being built now. Not finished yet.

Licence Subscription Monthly or annual, priced generously. Cancel at period end, and your data stays yours and stays exportable.

Download unavailable yet Documentation

Where this stands

Everything in the catalogue is still in development. You can read what each application does, what it needs, and what it cannot do yet. Nothing here is open to download or buy. Leave your email and you will hear from us the day that changes, once and about that only.

What it does

Talk with a voice model running on your own hardware.

Your model, your machine Point it at the computer running your voice model and press Link. Your voice goes to that model and its spoken reply plays back.

Three ways to take a turn Always listening, hold to talk with a button or the Space bar, or wait quietly for a wake phrase.

Cut in any time Start talking while it is speaking and it stops to listen.

A written record of every turn Everything said on both sides is written down as you go, and you can type a turn instead of speaking.

Tune it to the room Choose the voice, give it standing instructions, and set how loud speech has to be before it counts.

Known limit

You need your own computer running the voice model, with a large NVIDIA graphics card.

We list these because you will find them anyway. A limitation you learn about after paying is a defect in how we sold it to you.

Requirements

System A current desktop browser with a microphone

A computer with a large NVIDIA graphics card to run the model

Documentation

Installation, requirements, configuration and licensing for Beyond Voice Chat.

Read the docs

Release history

Every release, what changed in it, and the hash of exactly what shipped.

See releases