voicetranslator.app · English pages · updated 2026-09-13
Free tools, and numbers we measured ourselves.
Voice Translator is one product for people who understand the language but freeze when it is their turn: what they say, you read in Chinese; you speak Chinese and it shows them one large line in their language, numbers confirmed first. Three postures — Listen (they talk, I read), Talk (back and forth, on my behalf), Present (I talk, it draws) — on two devices, a Mac and an iPhone. Speech carries meaning, not facts — so dates, amounts and names are confirmed on screen first, and the app itself makes no sound. Building it produced two things worth publishing in English: a handful of small tools that run entirely in your browser, and first-party measurements of how fast live captions actually arrive. Both are on this page. The product pages themselves are written in Chinese, because that is who the apps are for.
Free tools — no upload, no sign-up
Each of these runs as JavaScript in your tab. Nothing you paste reaches a server, because there is no server to reach: this site is static files.
Calculator
Speech time calculator
How long a script takes to say. Measures your own pace with a stopwatch instead of asking you to guess, and converts a Chinese script to English speaking time at a measured 1.43 characters per word.
Converter
Subtitle converter
SRT to VTT, VTT to SRT, or plain text. Shifts timestamps, keeps two-line bilingual cues intact, and can split a bilingual file into one file per language.
Document
Bilingual document maker
Source and translation paired sentence by sentence into a printable sheet. Paste a back-translation and it highlights every word that no longer matches your original.
In other languages: スピーチ原稿の文字数と時間 · 対訳表メーカー（逆翻訳チェック） · 字數計算器 · 大數字換算（billion ↔ 億）
Measured, not claimed
Live translation is sold on adjectives. These are numbers from our own machines, each with the date and the method next to it: caption latency (Chinese) and how accurate Apple's on-device recognition and translation are, 11 languages to and from Chinese.
0.71 s
English words on screen, p50, in a real 31-minute meeting (2026-09-03)
2.81 s
First Chinese character of the translation, same meeting — the number we are least happy with
≈168 ms
Finish speaking to the translation on screen, fully on-device, Ukrainian ↔ Chinese (2026-09-05 run, minus the speech-output step the app no longer has)
1.43
Chinese characters per English word, across 1,052 aligned sentence pairs from real meetings
Guides built on the same first-hand work: recording system audio on a Mac (the ScreenCaptureKit failure nobody documents) , whether translation earbuds are worth buying, and what translated captions in Teams, Zoom and Meet require.
One product, three postures
A posture is defined by who the screen is for and whether it speaks — not by the venue. Switching between them does not end the session.
| Posture | Whose screen, what it shows | Mac | iPhone |
|---|---|---|---|
| Listen | They talk, I read. The screen is yours and it stays silent: what they said becomes Chinese captions, and nothing else is on screen. | Captures the meeting app's system audio; the captions float over Zoom | Lies on the table and listens to Ukrainian |
| Talk | Back and forth, always on my behalf. You speak Chinese; it shows them one large line in their language and makes no sound. Numbers, dates, amounts and names are confirmed on screen before the other side sees them. The other side does not press anything. | Listens to the person next to you through the microphone; English, French, German, Spanish, Portuguese, Italian, Japanese or Korean | One phone between two people; Ukrainian ↔ Chinese, on-device |
| Present | I talk, it draws. Each stretch of speech adds one step to a board — a flow, a tree, a list, or a note when it cannot tell — labelled in the other side's language. Undo the last step if it drew the wrong thing. | A board window you share into the call | The top half of the screen turns to face the other side |
Release status for both devices is on the download page (in Chinese).
Why the product pages are in Chinese
Because the products are built for one specific person: a native Chinese speaker who has to work in English and is tired of understanding the meeting perfectly while saying nothing in it. Writing their product pages in English would make them worse, not wider.
What generalises is on this page — the tools, the measurements, the guides. If you build something in this area and want the raw numbers or the method, the contact address is in the footer, and the author is on GitHub. The Chinese pages start at voicetranslator.app.