Voice as an interface, not a feature

Most of the voice products we've tried treat voice like a novelty button. There's a chat widget in the corner, and next to the text input there's a little microphone. You can dictate your message, and the product transcribes it. Maybe it even speaks the response back.

That's not voice. That's a chat UI with a microphone taped on. It still requires you to think in the shape of a message — compose a sentence, wait for a reply, compose another. The flow is fundamentally textual. The voice is cosmetic.

Real voice interfaces work differently. You speak in half-sentences. You interrupt. You change your mind mid-thought. You say "hold on" and the other side actually stops talking. You ask three questions in a row and expect three answers in order. The system has to listen, not just transcribe.

We designed Seny for that. The agent is interruptable — if you start talking while it's mid-sentence, it stops and listens. It handles overlapping turns, back-channeling ("mhm", "yeah"), and thinking pauses. It doesn't wait for you to press a button. It doesn't transcribe first and then respond. It's a real conversation.

The chat UI is there as a fallback for when you can't talk — in a noisy coffee shop, in a meeting, in an open-plan office. But the default is voice. That's the product.

We think this distinction matters because it changes what visitors feel when they use the site. A chat widget feels like paperwork. A voice agent feels like talking to someone.

— The Seny team