Shuovo

“Shuo” is 说 (shuō, to speak); “vo” is 我 (wǒ, me) — 说我, “speak my mind.”

For thoughts that take a few minutes, not a few seconds.

Shuovo is a voice keyboard for working something out loud — explaining a bug to an AI assistant, talking through a product decision, drafting a post in Chinese with the English terms left in. Talk, stop to think, check something, change your mind, keep going. When you stop, the whole session comes back as one clean draft — in the text field you started in, on Android or in Chrome.

Five weeks of the founder’s own use: 509 sessions, 17 a day. One in three ran past a minute; the longest ran 19. Four in ten seconds were silence. One in three was in Chinese, or both languages at once.

A real session, lightly edited

0:00“Let’s think about how to make this better. The part I’m least happy with is the accuracy — sometimes it invents words, and sometimes it’s wrong in a way that doesn’t fit the context at all.”
0:46“One idea — and this might not be the best one, I’m just outlining it as a first step — is to run the audio through several speech models and compare. With one source you can’t tell what’s wrong.”
3:11“So the target has to be single digits. Maybe one error in a thousand. Right now it’s about…”
3:39Fourteen seconds of silence. Trying to remember a number.
3:53“…three to five percent, if I remember correctly. Slightly higher for the other one. But my main point is each provider has its own data.”
5:40“Actually, there’s a different route: let people submit a sample of how they say a word, and learn from that.”
7:47“About three seconds. Five for a long recording. That’s the budget.”
8:42Stop. About 900 words spoken. One structured draft lands in the AI chat that was already open.
afterA word or two fixed with the keyboard. Send.

That was one of 509 sessions in five weeks. Most were under a minute — a follow-up, an instruction, a question. One in three ran past a minute, and those averaged close to three. Four in ten seconds of them were silence, which is where the thinking happened. Ordinary dictation ends the input at the first of those pauses. Shuovo treats it as the middle.

Where those sessions landed: mostly AI chats — debugging a laptop that wouldn’t boot, shaping a report-generation product, weighing how to structure a partnership — and the rest in WeChat messages, posts, notes to self, and a six-minute restaurant cost note dictated in Chinese with half-minute pauses to count.

Is this for you?

If you already know what you want to say, any dictation tool will type it for you faster than your thumbs. Shuovo is for the other kind of thought — the four-minute explanation of a bug to an AI assistant, the post drafted in Chinese with the English terms left in place, the plan you can’t get out until you’ve talked through it, checked two things, and changed your mind once. It isn’t quicker at a sentence. It’s built for the minutes.

Start with 20 seconds

Losing a sentence is annoying. Losing eight minutes of thinking isn’t. Try Shuovo here for twenty seconds, see what comes back, and decide whether it earns a longer session.

Ready when you are.

0:00

Text to edit

In the app you’d select text and speak. Here, your spoken instruction is applied to this sample — feel free to change it.

Try saying

Okay, so the part I’m not happy with is the accuracy. One idea — and it might not be the best one — is to run the audio through several models and compare. Actually, let me think about what that costs first…

Result

Show raw transcription

This demo runs on separate, smaller infrastructure than the app, and takes short clips only.

The Android app and Chrome extension are free to install. Voice is a subscription — 14 days free, then $15/month — with no session cap.

Why use Shuovo

Open another app. The session is still running.

Start talking, then go and check — the email, the spreadsheet, the source, the other AI chat. Come back and pick up mid-thought. On Android the keyboard keeps recording while you’re in the other app; in Chrome it keeps recording while you switch tabs. One session, one draft, however many places you had to look.

Nineteen minutes you can’t lose to a glitch

There’s no session cap — the longest of the founder’s 509 sessions ran 19 minutes, in Chinese, about 1,700 words. Your audio is saved as you speak, before transcription starts, so a timeout, a dropped connection or an app crash doesn’t take the recording with it. A long session is handled as a long session, not as a short one that got lucky.

Spend less time fixing the result

Shuovo is tuned for the time between finishing speaking and having text you’d actually send — not for shaving the last second off transcription. Over an eight-minute session that means fewer wrong names, fewer mangled numbers, and less of the draft you have to re-read.

A draft, not a transcript

Filler words, restarts and the three times you said “actually, no” are cleaned up. What lands has paragraphs, punctuation and the structure you were reaching for out loud.

The keyboard is the last mile

It’s a keyboard so the draft can go into any text field on the device, and so the last few corrections are one tap away instead of a copy-paste round trip. Select a phrase and say the change, or delete a stray word or a whole sentence in one tap.

中文 + English in the same sentence

No mode button, no language toggle — say it the way you’d say it to a colleague who speaks both: “这个 design partner 愿意先给反馈” comes back exactly like that. One in three of the founder’s sessions was Chinese or both languages, and the English terms stayed English. On the Android keyboard, tone-assisted input fixes the occasional mis-heard character in a tap or two.

Every session stays yours

A recording that didn’t get transcribed — the connection dropped, the tab closed — is kept, and you can retry it. On Android you can also review, copy, export or search any past session. Nothing is stranded in an app you’ve already closed.

Free app — voice has a 14-day free trial

The keyboard is free forever. Voice is $15 / month after the trial — sign up here to unlock it. Cancel anytime.

What you’ll need

An email address — we send a sign-in link, no password.
A subscription after the trial — $15 / month, cancel anytime. Checkout asks for a card up front, but nothing is charged until the 14-day trial ends — cancel before then and you pay nothing.

Shuovo exists because I spent years thinking in two languages, six hours a day, with wrists that had stopped forgiving the keyboard — and every dictation tool I tried assumed I knew what I wanted to say before I said it.