Demo — Voice AI
Voice assistant.
Ask a question about Friendify out loud. It listens, answers from this site and speaks, with four open models running in this page. Your voice never leaves your device.
Checking what this device can run…
Recorded run
Recorded on 28 October 2026, on a desktop computer with an NVIDIA RTX 4060 Ti graphics card, in Chrome. The timings are that machine’s.
You
What does Friendify build?
Assistant
Friendify builds products that make technology feel more human, and they are designed to last.
- Heard
- 2.2 s
- Understood
- 0.4 s
- First words
- 1.7 s
- Spoke
- 3.1 s
You
Do you make voice assistants for phone lines?
Assistant
Yes, we make voice assistants for phone lines and apps.
- Heard
- 3.2 s
- Understood
- 0.3 s
- First words
- 0.5 s
- Spoke
- 0.9 s
From: Voice AI
You
How long do you keep my contact message?
Assistant
We keep your contact message for up to 24 months after you send it, then delete it.
- Heard
- 3.2 s
- Understood
- 0.6 s
- First words
- 0.9 s
- Spoke
- 2.3 s
How it works
Listen, think, speak, and stop when you talk.
Four open models in one page, passing the turn from one to the next. This is the run behind the recorded example, step by step.
You say
Do you make voice assistants for phone lines?
Step 1
Listen
It notices when you start and stop talking. The microphone stays in the page.
3.2 seconds of speech
Loudness detector
Step 2
Write it down
A small open speech model turns your voice into text.
“Do you make voice assistants for phone lines?”
Moonshine · 67 MB · 0.3 s
Step 3
Find what the site says
The nearest passages of this site, as in Ask the site.
Voice AI · phone and in-app voice agents
bge-small · 35 MB
Step 4
Answer from them
A small language model writes one or two sentences, from those passages only.
First words after 0.5 s
Qwen3 0.6B · 579 MB
Step 5
Speak
The first sentence is spoken while the rest is still being written.
First sound after 0.9 s
Kokoro · 335 MB
It says
“Yes, we make voice assistants for phone lines and apps.”
- from your last word to its first, on a desktop graphics card
- 0.9 sfrom your last word to its first, on a desktop graphics card
- 1.0 GB with the engine that runs them, downloaded once
- 4 models1.0 GB with the engine that runs them, downloaded once
- of your voice sent anywhere
- 0 bytesof your voice sent anywhere
Good to know
A laptop’s worth of models.
- The four models come to a large download and need a graphics chip the browser can use (WebGPU: Chrome, Edge, or Safari 26 on a laptop or desktop). On phones and other browsers you see a recorded run instead.
- Hands-free works best with headphones: the browser cancels most of the assistant’s own voice, but not all of it in a loud room.
- The language model is small, so it keeps to short answers about what this site says, and says when it doesn’t know. The assistants we build for clients use larger models, their own knowledge, and a person to hand over to.
- The microphone is used only while the demo listens. Nothing you say is recorded, sent or kept.
About this demo
Small on purpose.
The demos here download 50 MB to 1.0 GB: small enough to fit into a web page and run on your own device, with nothing sent to us or to an AI company. Your browser keeps the files, so a second visit starts from that copy.
They do a great deal for their size, and they make mistakes a larger model wouldn’t. What we build for clients uses larger models where they’re needed, the client’s own data and words, testing before launch, and a person to hand over to.
The files stay in your browser until you remove them, or the browser needs the space. Removing them here clears every demo’s models from this device; a demo downloads its own again if you load it.