speech in · sight out
You talk. It shows.
Speech in, sight out — the next interface after chat.
a diagram of the idea — not a product
the idea
The chat bubble is a bottleneck in both directions
in
Typing is the slowest way to tell an assistant anything. It is also the only way most assistants offer.
out
A paragraph of prose is the weakest way to receive a plan, a list, a comparison, a picture.
The movie Her is the inspiration, taken literally on one axis.
You talk to your assistant the way you talk to a person. And the answer does not come back as a wall of text read into your ear — it appears on a window that belongs to you: lists, plans, tables, images, controls. The voice carries the conversation, the display carries the substance.
Speech in. Sight out.
Think of the input half as dictation done right: always there, low latency, no app to switch to. The output half is the part dictation has never had — a surface the assistant draws on while you speak.
what already exists
Not shipped, but not floating either
voice2vision is built on meclaw, an open substrate that already carries three pieces of it — and is missing the one that matters most.
The window
A screen is a channel, and a channel belongs to the person, not to an agent. Several agents hold views on the same window, and none can touch another's. One screen, one app today.
Sessions are already phone calls
A conversation is modelled on a phone call: it opens, it runs on an idle clock, it ends. The speech-shaped session model is in the substrate before speech is.
Images travel inward
An assistant can already be shown a picture. Vision in works. The idea adds the mirror direction — vision out.
The entire voice half
Capture, transcription, latency budget, barge-in, and the judgement of what deserves the screen rather than the voice. None of it exists. It is a design round of its own before it is even a task.
status
voice2vision is an idea in the open. Nothing here ships today.
There is no product, no waiting list and no date. This page exists because the idea is worth writing down before it is worth installing — and because the substrate underneath it is public, so the story can be followed as it happens.
If you want to watch it get built, the work happens here: