--- title: "On the screen, in your notes, or saved for later · VoiceMe" description: "Floating subtitles, words typed into your notes, a transcript you keep, and the things it does on its own." url: "https://royalsoftworks.com/products/voiceme/ways-to-read-it/" source: "https://royalsoftworks.com" format: "markdown" note: "Markdown rendering of the HTML page at `url`. Same content, same canonical URL." --- VoiceMe · 2 of 5 # On the screen, in your notes, or saved for later. Floating subtitles, words typed into your notes, a transcript you keep, and the things it does on its own. Three ways to get the words ## On the screen, in your notes, or saved for later. Pick one, or run all three at once. They are the same words arriving in three different places, and none of them is switched on until you say so. Subtitles on screen Typed into your notes Kept as a transcript ### A window that floats over whatever you are doing. The subtitle window sits above every other window and follows you between desktops and into full-screen video. Make it see-through so it looks painted onto the screen, give it a solid panel if that reads better, or fill it with a flat green that streaming software keys straight out — which is how you put live captions on a broadcast without paying a captioning service. - See-through, solid, or a flat colour for OBS and vMix - Borderless and click-through, so it never gets in your way - Can be hidden from screen recording and share-your-screen calls - Your font, size, colour, outline or shadow, line count and fade-out and marketing signs off on Friday morning See-through Green screen Solid panel The technical detail - **Window** — Always-on-top, borderless, click-through, visible on all workspaces and over full-screen video - **Backgrounds** — Transparent, solid, or a flat chroma-key fill for keying in OBS / vMix - **Capture** — Optional exclusion from screen capture and screen sharing ### The words land in whatever you are writing in. Put the cursor in a document, a chat box or a ticket and VoiceMe types what it hears into it. Because that means letting a program type into other people's documents, it is deliberately awkward to switch on: you pick a mode, then you separately arm it, and it disarms itself every time the app restarts. One keyboard shortcut turns it off from anywhere, and you can name applications it must never type into. - Types into any application that has a text field - Two switches and a panic shortcut before a single character is typed - A block-list for apps it must never type into, such as your password manager - Arming does not survive a restart — it is off again every time you open it armed · ctrl+alt+T stops it meeting-notes.md heard “so the deadline moves to the 14th” typing → meeting-notes.md heard “and marketing signs off on Friday” typing → meeting-notes.md The technical detail - **Method** — Clipboard paste or synthesised keystrokes — SendKeys on Windows, xdotool / wtype on Linux. No native modules - **Safety** — Two-stage arming, a global disarm hotkey, a per-application block-list, and disarm on restart ### Everything it hears builds up a record you can read back. Whether or not the subtitles are on screen, the words accumulate into a transcript you can scroll, search and save. Keep the lecture, the podcast or the recorded meeting, and export it in the shape whatever comes next wants to read: plain text for a person, Markdown for a document, one line of JSON per sentence for a script. - Scroll back and search everything heard this session - Export as plain text, Markdown, or one line of JSON per sentence - Timings are kept with each line - It is a file on your disk, not a page on somebody's website 00:04:12 the deadline moves to the fourteenth 00:04:19 and marketing signs off on Friday 00:04:26 we can push the rest to the next review TXT Markdown JSONL The technical detail - **Export** — TXT, Markdown, JSONL. There is no subtitle-file (SRT) writer - **Storage** — Written where you choose, on your own machine. Nothing is synced While you are not looking ## The four things it does on its own. ### Works with the internet unplugged Once the model is on your disk there is nothing left to fetch. No account, no sign-in, no key to paste, no server to be down. ### Around 99 languages It can recognise speech in most of the world's major languages, and it can translate what it hears into English as it goes. ### The free version has one limit Only the floating subtitle window is timed — an hour, a five-minute pause, then another hour. Listening, transcribing, typing and saving are never limited. ### Updates itself, quietly It checks once, twenty seconds after you open it, and never again while you are using it — so it cannot interrupt an afternoon of captioning. [Previous What it listens to](https://royalsoftworks.com/products/voiceme/how-it-listens/) [Next What people use it for](https://royalsoftworks.com/products/voiceme/uses/) [Back to VoiceMe overview](https://royalsoftworks.com/products/voiceme/)