Web Speech API (SpeechRecognition) is required. Try Chrome, Edge, or Safari 14.1+/iOS 14.5+ (sends audio to Apple's own servers, a third engine distinct from Chrome/Google or Edge/Azure). Firefox doesn't implement it.
Your device's speech engine may not support this language — check your phone's system language/voice-input settings (common on some Android OEM builds, e.g. MIUI).
Runs in the background — it won't slow the live transcript down (translations appear a moment later). Uses one free, key-less service at a time (Google or MyMemory). No automatic fallback: if a line can't be translated it's shown in red — switch service here and later lines retry with the new one.
Choosing a folder to save into needs desktop Chrome / Edge — elsewhere recordings stay in this browser.
Recordings stay in this browser. Pick a folder in ⚙️ Settings → 📁 Choose folder… (desktop Chrome/Edge) to save real files.
No recordings yet — tick ⏺️ Record audio on the Live tab, then ▶ Start.
Open a recording someone sent you
Got a recording from a colleague? Pick the files they sent — an audio file (.webm, .m4a, .mp3…), and, if they included them, the matching .txt transcript and .cues.json. It plays right here with the transcript. Everything stays on your device — nothing is uploaded, and you don't need to choose a save folder for this.
🔄 Version & update ●
Current version: …. Tap the button to check for a newer version and refresh.
📐 Layout
You're on the mobile layout. 🖥️ Switch to the desktop layout — bigger fixed panes, drag-to-resize split. Best on a computer; on a phone this mobile layout is recommended.
🎙️ Recognition
Recognition language (日本語 / English / 中文) is chosen on the 🎙️ Live tab and remembered for next time.
ℹ️ When captions feel slow or freeze on a phone
Phone speech engines get slower the longer one session runs, so the app quietly swaps in a fresh recognizer on a timer. Auto is 180s on phones, 50s on desktop. If captions on your phone lag or stall, pick a shorter value (try 45–60) — it recovers sooner, at the cost of a <1s gap on each swap. A longer value means fewer gaps. Takes effect the next time you press ▶ Start.
Auto-adjust: while recording, if captions go silent for 20s+ the app shortens this interval one notch (recovers sooner, floor 30s); after a few minutes of steady captions it lengthens it one notch again (fewer gaps, up to the Auto value). It keeps whatever value it lands on.
Caption smoothing controls how often the still-being-typed line is redrawn while you speak fast. Higher = steadier, less flicker, but the in-progress line updates a little less often. Finished sentences always appear immediately. Takes effect the next time you press ▶ Start.
🕐 Recent caption restarts (0)
Every time the recognizer swaps in a fresh instance (proactive Caption refresh, or recovering from a stall), it's logged here with the gap since the previous one — so you can see whether a "stopped then resumed" moment was this normal recycling, not a real problem.
📺 Display
ℹ️ About these settings
Font: System sans is the OS default sans-serif (best for CJK); Serif / Mincho has contrasting strokes; Rounded is a soft sans; Monospace is fixed-width. Text size: S / M / L1 / L2 are quick presets; pick Custom to type an exact px value. Size and font apply to both the transcript and the live-translation panel. Width: Normal is a centered column; Wider is about 1100px; Full width uses the whole screen; Custom sets a max width in px. The px box next to Width / Text size is editable only when the dropdown is set to Custom — otherwise it just shows the preset's value. Keep last N lines drops older lines once the box grows past N (and trims that recording's saved .txt the same way); 0 keeps everything. ↺ Reset restores defaults.
⚙️ Advanced & less-used
🎤 Microphone access
If the page says the microphone is blocked / disabled, use this to check it and to see how to turn it back on. A web page can't un-block the mic or open the browser's settings by itself — that's a browser security rule — but the button below re-asks for permission, and if the browser has hard-blocked it the steps to fix it appear.
Permission state: checking…
🚫 It's blocked — how to allow it (Chrome / Edge on Windows)
- Look at the left end of the address bar for a 🔒 lock or tune / sliders icon → click it.
- Click Site settings (Chrome) or Permissions for this site (Edge).
- Find Microphone → set it to Allow.
- Come back to this tab and reload the page, then press 🎤 Test microphone again.
- Still blocked? Check Windows itself: Settings → Privacy & security → Microphone → turn on Microphone access and Let desktop apps access your microphone.
⏺️ Recording audio
The ⏺️ Record audio switch is on the 🎙️ Live tab, before ▶ Start — turn it on before you press Start.
ℹ️ How saving works
⏹ Stop just stops listening — it never saves and never asks. To keep a segment, press 💾 Save (tick 🔊 Audio and/or 📝 Text): with Audio you get audio + transcript as one History entry; Text alone saves the transcript .txt.
If you stop with recording on and don't Save, that take is kept in memory only — tick 🔊 Audio + 💾 Save to file it into History; it's dropped when you start a new session or reload.
Recording opens a second microphone stream (MediaRecorder, WebM/Opus). Nothing is uploaded anywhere.
🖥️ Transcribe the sound your computer is playing
Captions come from the browser's speech recognition, which only hears your current system recording device (microphone). A web page can't grab the speaker output with a single button — that's a browser security limit. To transcribe what the computer is playing (a video, a call, an online class), route the playback into a recording device in Windows. Two ways:
Or use the desktop version — it captures system audio directly (WASAPI loopback), no routing needed, and transcribes on your own machine (offline, private). Ask the maysuns.uk team about the desktop live-caption app.
Option A — Stereo Mix (no software to install; works on most desktops / some laptops)
- Right-click the volume icon in the taskbar → Sound settings.
- Scroll to the bottom → More sound settings (the classic Sound panel).
- Go to the Recording tab → right-click an empty area → tick Show Disabled Devices.
- Find Stereo Mix → right-click → Enable → right-click again → Set as Default Device.
- Come back here, press ▶ Start, and when the browser asks for a microphone choose Stereo Mix.
- When done, set your normal microphone back as the default recording device.
If there's no "Stereo Mix" in the Recording tab, your sound driver doesn't provide it — use Option B.
Option B — virtual audio cable, VB-CABLE (works on any machine; installs one free tool)
- Download and install VB-CABLE (vb-audio.com, free) → restart the computer.
- Sound settings → set the Output device to CABLE Input (no sound from the PC now is expected).
- Sound settings → set the Input device to CABLE Output.
- Come back here, press ▶ Start, and choose CABLE Output as the microphone.
- To also hear it yourself: Sound settings → CABLE Output properties → Listen to this device → play through your headphones / speakers.
- When done, set Output and Input back to your normal devices.
⏺️ Record audio uses the same device and captures it raw (no echo cancellation), so loopback sound isn't treated as an echo and removed.
📁 Save location
Not set — recordings and saved transcripts download to your browser's default Downloads folder.
Your browser doesn't support choosing a folder (needs desktop Chrome or Edge — not Android Chrome, Firefox, or Safari/iOS as of 2026-08). Files download to your default Downloads folder instead.
ℹ️ Where do my files go? (full path)
When you pick a folder, recordings (Record0001-20260828-1530-1532.webm) + their transcript (.txt, same name) + a small .cues.json for synced playback are written straight into that folder — real files you can open, play, or send.
A web page is not allowed by the browser to know or show the folder's full path (e.g. C:\Users\you\Recordings) — that's a privacy rule in every browser, not a missing feature. It also can't open the folder in your file manager for you. To find the files: open the folder you picked in Windows Explorer / Finder yourself; the file names above tell you exactly which is which (date + start/end time). If you need the app to show and open a real path, that requires the desktop version, not a web page.
🌐 Translation service
Turn a running translation on with the 💬 Translate button on the Live tab (on the Save row). The target language and the service (Google / MyMemory, both free) are set in the panel that appears under the transcript when it's on.
📥 Add to desktop / home screen
- Puts an app icon on your desktop or phone home screen.
- Opens in its own window — no browser address bar.
- Still runs in your browser. Settings and saved recordings don't change.
On desktop Chrome / Edge this installs the app directly. On a phone, follow the steps below.
ℹ️ Install steps by device
- Desktop Chrome / Edge:
click 📥 Install on this device → Install.
(Or use the install icon in the address bar.) - iPhone / iPad — Safari:
tap Share (square with an up arrow)
→ Add to Home Screen → Add. - Android — Chrome:
tap the menu ⋮ (top right)
→ Add to Home screen / Install app → Add. - Firefox:
can't install as an app — use a bookmark.
👥 Multi-speaker mode
🚧 Not implemented yet. Reliable speaker separation needs real voiceprint analysis, which the browser speech engine can't provide — the earlier volume/pitch guess wasn't dependable enough to keep, so it's been removed for now.
What this is
- Live speech-to-text for meetings and notes.
- Speak Japanese, English, or Chinese — it types what it hears.
- No translation of the transcript itself. Turn on 💬 Live translation under the transcript for a running translation, or press 💾 Save and translate the text elsewhere.
- Runs in your browser. Nothing is sent to our server.
How to use
- Pick 日本語 / English / 中文 on the Live tab.
- (Optional) tick ⏺️ Record audio (before Start) — or あ Furigana for Japanese readings.
- Tap ▶ Start and allow the microphone.
- Talk — text appears as you go.
- ⏸ Pause to hold, ▶ Resume to carry on (same session).
- ⏹ Stop when done. Nothing is saved yet — the text stays in the box.
- Press 💾 Save to keep it. For a translation, turn on 💬 Translate below the transcript.
The buttons
- ▶ Start / ⏹ Stop — one button. Start begins listening; Stop stops listening. Stop leaves the text in the box and never saves and never asks.
- ⏸ Pause / ▶ Resume — a temporary hold that keeps the same session (and the same recording) going. Different from Stop, which ends the session.
- 💾 Save box (under the transcript) — the only Save. Tick 📝 Text and/or 🔊 Audio, then Save. Text saves the transcript
.txt; Audio (available when a take is waiting) files the recording into 📜 History with its transcript. Greyed while recognition is running — 💾 Save, 🌐 Translate and 🗑️ Clear all become usable once you Stop or Pause. - 🗑️ Clear (right end of the Save row) — empties the text box for a fresh session. Doesn't touch History (an ↩️ Undo shows for a few seconds).
- ⏺️ Record audio and あ Furigana (top row, before Start) — turn on before you press Start. Furigana's first use downloads a ~17 MB Japanese dictionary once, then works offline.
- An unsaved recording (Stop with Record on, didn't Save) is dropped when you start a new session or reload — tick 🔊 Audio + 💾 Save to keep it.
- When recognition is stopped or paused the transcript box is editable — type corrections in, and 💾 Save (📝 Text) keeps them.
Why captions are sometimes slow or wrong on a phone
- How accurate the recognition is comes from the phone's own speech service (Google on Android Chrome, Apple on iOS) — no setting in this app makes recognition itself more accurate. Speaking clearly, close to the mic, in a quiet room helps most.
- Android: use Chrome — a built-in browser (Samsung Internet, a Mi browser, the in-app browser of another app) may use a weaker engine or none at all.
- Install it as an app (Settings → ⚙️ Advanced → 📥 Add to desktop / home screen). A standalone window keeps running more reliably than a browser tab you switch away from.
- Pick one language and stay on it for the whole session — switching mid-session restarts the recognizer and drops a moment of audio.
- A wired earbud mic beats the built-in mic in a noisy room. Don't use a Bluetooth headset as the mic.
- Captions lagging or frozen? Settings → 🎙️ Recognition → Caption refresh → 45–60s, or tick Auto-adjust.
- In-progress line flickers while you talk fast? Settings → 🎙️ Recognition → Caption smoothing → Extra or Maximum. Finished sentences still appear instantly.
Display & history
- Width and 📺 TV mode (Settings tab) are two independent switches — Full-width just widens the column; TV mode is a big-text dark reading look. Turning TV mode off puts everything straight back. Text size (a px number) and font apply to both the transcript and the translation panel, at the same size.
- 📜 History — the top row is the multi-file action bar: Applies to: 🔊 Audio / 📝 Text + ⬇️ Download / 📤 Share / 📦 Move… (grouped — they're the same kind of thing). They act on the recordings you've ticked, or on the one you've opened if none are ticked, honouring the 🔊/📝 switch. 🗑️ Delete selected appears when rows are ticked. Drag the bottom edge of the list to resize it. Pick a recording to open its detail: Export subtitle (.srt/.vtt), Save a copy (🔊/📝, name, 📄 Save as…), then the Transcript box (edit + Save). A recording with timings also has a resizable follow-along view.
Browsers
- Speech recognition: Chrome, Edge, and Safari 14.1+ / iOS 14.5+ (each sends audio to its own maker's servers).
- Firefox: not supported.
- Saving into a folder + browsing it: desktop Chrome / Edge only. Elsewhere, recordings are kept in this browser and still show on History.
What your browser can do right now:
On iPhone / iPad
- iOS makes every browser (Safari, Chrome, Edge…) use Apple's speech engine — it's weaker than desktop Chrome/Edge and can stop after roughly a minute.
- Use the mobile layout (this one) and keep sessions short — press ⏹ then ▶ every few minutes.
- Captions frozen? Settings → 🎙️ Recognition → Caption refresh → pick 45–60s, or tick Auto-adjust.
- A wired earbud mic (USB-C / Lightning EarPods) or a wireless lav with a plug-in receiver (DJI Mic Mini, etc.) beats the built-in mic. Don't use a Bluetooth headset as the mic — iOS drops it to call quality.
Learning from video
- 🎬 Bilingual Subtitle Editor — load a
.srt/.vtt, translate it line by line, edit, and export your own bilingual subtitles. All in your browser, for private study.
About / contact
- A free tool by the team at maysuns.uk — more free tools there.
- Feedback: get in touch about Live Transcript.