00:00:00,000Changelog

Changelog

What’s new in each version of Live Transcript, newest first. The app shows the same notes when it updates.

Live Transcript 1.5.0

· build 134

Live Transcript now speaks German, Japanese, Vietnamese, French, Spanish, Italian and Polish.

New

  • The whole app is translated: menus, Settings, History, the notch, the setup wizard, licensing and the macOS permission prompts. It follows your Mac's language.
  • Choose a different language just for Live Transcript in Settings › Overview › Language. It's the same setting as System Settings › General › Language & Region › Applications, so the two always agree.
  • Prices stay in US dollars, written the way your language writes numbers.
  • New transcripts label you and the other side in your language ("You" and "Them" in English). Earlier transcripts keep their labels.

Live Transcript 1.4.1

· build 127

Claude without an API key works again.

Changed

  • During the free trial, the menu, Overview and the notch show how many days are left, with a Buy button next to it. Your trial and license stay exactly as they were.

Fixed

  • "Claude, through the installed claude tool" in Settings › AI couldn't find Claude Code on some Macs (for example with iTerm2's shell integration) and could run an old copy that no longer starts. It now finds the copy you use in Terminal and passes over copies that don't work.

Live Transcript 1.4.0

· build 123

Tell apart the people sharing your microphone.

New

  • Speaker labels on your side too: turn on "Label speakers on my microphone" in Settings › Sources and several people around one mic become Speaker 1, Speaker 2 and so on, just like the other side of a call. It also covers any extra input devices you listen to. It uses the same speaker model as the other side, so there's nothing new to download if you already label speakers. It's off until you turn it on, and like before, labels belong to the session and no voiceprints are kept.
  • When both sides have several people, names say which side they're on ("You · Speaker 1", "Them · Speaker 1"). Click any label to rename it.

Live Transcript 1.3.0

· build 121

Share the notch with more apps.

New

  • Sharing the notch is open to other notch apps that support it, not only Polytopic's: any app signed with a Developer ID or from the Mac App Store can take turns at the notch with Live Transcript, in the sharing style you pick. To keep one out, right-click it in the list in Settings › Captions.

Changed

  • With the Switcher style, the dots for each app sit in a row of their own under the notch, so they no longer crowd the controls in the open pill.
  • Live Transcript 1.3.0 shares the notch with Tattle 1.1.0 or later (1.1.1 for the new dots row in Tattle too). Update both, or each falls back to making room for the other.

Live Transcript 1.2.0

· build 117

Know who said what on the other side, a guided setup, and less work while nobody talks.

New

  • Speaker labels. When several people talk on the other side of a call or video, their lines are labelled Speaker 1, Speaker 2 and so on, in the order they are first heard. A line is split where the speaker changes. Your microphone keeps its own label.
  • Click a speaker's name to rename them for that session, in History, the floating captions or the notch.
  • A second, slower pass tidies the labels while you listen, about half a minute behind, and settles them when the session ends.
  • Speaker labels are off until you turn them on in Settings › Sources › Speakers or in Setup. They need a one-time download of about 154 MB, and the tidy-up pass another 154 MB. Everything runs on your Mac; no audio leaves it, and no voice is remembered from one session to the next.
  • A new Setup guide walks through sources and permissions, the speech engine and language (with each download's size), speaker labels, and the notch and captions. Open it any time from the menu bar's Setup….

Changed

  • While nobody is talking, the speech engine rests and picks up again with the first word, so long sessions use less battery. Nothing is missed: recording and the level meter keep going.

Fixed

  • Choosing Download Now for a Whisper model and then starting to listen no longer downloads the same model twice.

Live Transcript 1.1.0

· build 91

Change things while you listen, take the whole conversation’s audio with you as one file, and no more echo from your speakers.

New

  • Changes apply while you’re listening, in the same session. Turn the microphone, your Mac’s sound or a single app on or off, pick another microphone, or rename a speaker, and the session carries on. Lines already written stay where they are.
  • Switch the speech engine or the language mid-session. Captions keep coming from the current engine while the new one gets ready (it may need to download first, shown at the notch), then it takes over.
  • Record Audio applies at once too. While it’s off, the recording keeps time with silence, so lines still play back at the right moment.
  • A speech engine picker in the menu bar menu and in the notch: hover over the notch and click the engine’s name.
  • One audio file for the whole conversation, with every speaker lined up in time. Adding a session to Apple Notes now attaches this one file instead of one per speaker, and History has a new Export › Audio… to save it anywhere. Sessions recorded with earlier versions work too.

Changed

  • The recent transcript in the notch holds still while you scroll back to read it, and shows how many new lines came in meanwhile. It used to jump back to the newest line.
  • Start New Session (Apply Changes) is gone from the menu bar menu, since changes no longer need a new session. New Session is still there when you want a fresh start.

Fixed

  • The microphone keeps working while another app is using it, such as a FaceTime or Zoom call. Your side of the conversation could go quiet before.
  • When a call plays through your Mac’s speakers, the microphone hears the other side as well. Live Transcript now takes that echo out by itself, so their words no longer appear a second time as yours. With headphones nothing changes.
  • The Echo cancellation switch in Settings › Sources is gone. While it was on, Live Transcript heard almost nothing of your Mac’s own sound. Echo is now removed automatically instead.
  • Whisper no longer writes lines nobody said, such as “Thank you.”, while the microphone is quiet.
  • The menu bar menu stays open while captions come in. It used to close each time new words arrived.

Live Transcript 1.0.2

A menu shortcut for recording, and a fix for Apple Notes.

New

  • Record Audio is now in the menu bar menu, under Microphone. It applies from the next session, the same as the switch in Settings › History.

Fixed

  • Sending a session to Apple Notes attaches each recording once. Notes was showing the same recording twice.

Live Transcript 1.0.1

A fix for the floating captions panel.

Fixed

  • Floating captions now appear when you turn them on after the app has started. In 1.0.0 the panel could stay empty unless it was already on at launch.
  • The floating panel’s grab handle and hover controls respond again.

Live Transcript 1.0.0

The first release of Live Transcript: live captions for everything your Mac plays, transcribed entirely on your Mac. It needs a Mac with Apple silicon running macOS 26 or later.

Captions at the notch, or in a floating panel

  • While someone is speaking, captions drop down from the notch with a level meter and a red dot showing it’s listening. When they stop, the notch tucks away.
  • Words appear lighter while they’re still being recognised and settle as they’re confirmed.
  • The floating glass panel shows the running transcript in a window you can move anywhere and pin above other windows.

What it hears

  • Everything your Mac plays, labelled “Them”.
  • Your microphone, if you turn it on, labelled “You”. With both on, a meeting reads like a script: who said what, in order.
  • It captures your Mac’s sound with macOS’s own Core Audio taps: no virtual audio device to install, and the sound keeps playing normally.

Speech engines

  • Apple Speech, or Whisper in three sizes: Low (145 MB), Medium (216 MB) and High (632 MB), each downloaded the first time you use it.
  • The engine wizard has you read a passage aloud once, runs every engine on that same recording, and shows how each one did with your voice.
  • More than one language: set Language to Auto-detect and Whisper works out the language of each phrase on its own; Apple Speech switches between the languages you list.

History and plain-language search

  • Every session is saved to History on your Mac, as cards grouped by day, each with a title, a one-line summary, the app that was playing and its topics.
  • Search the way you remember it: dates (“last week”), apps (“in FaceTime”) and kinds of listening (audiobook, video, meeting) are understood. Each result shows the passage that matched and why.
  • Ask a question and a short answer, generated on your Mac, appears above the results with numbered links to the moments it used.
  • Search, its index, and the titles, summaries, topics and answers stay on your Mac. With Apple Intelligence, they come from Apple’s on-device model; without it, search still works on words and meaning.

Two-way live translation

  • The translated line appears right under the captions as people speak, using Apple’s translation on your Mac.
  • With Language set to Auto-detect, it translates both ways between the first two languages you list, so both sides of a two-language conversation can follow; with one language, every line goes into the language you choose.

Sharing the notch

  • If another app is already using the notch, Live Transcript notices and makes room, adapting as that app launches, quits or pops up. Hold Option while hovering to leave the notch to the other app.
  • It never moves or hides other apps’ menu bar items, and needs no extra permissions to share the notch.

Export and Apple Notes

  • Export any session as Markdown, plain text, or SRT or VTT subtitles, or send it to Apple Notes with the recorded audio attached.

Summaries and questions, with the AI you choose

  • Ask a session for a summary (a short summary, key points and action items) or ask it a question.
  • Choose Apple’s on-device model, Claude (through the installed claude tool or with an Anthropic API key), OpenRouter, or a custom server that speaks the OpenAI protocol, such as Ollama or LM Studio.
  • Only transcript text you ask about is sent, and only to the provider you picked; with Apple’s on-device model, nothing leaves your Mac. Audio is never sent.

Licensing

  • A license is a key: no account, no name. It’s checked on your Mac, so it doesn’t need a connection to keep working day to day.
  • Personal covers 3 Macs and Business 10. Unregister a Mac in the app or on the account page and the install is free for another Mac.
  • You buy a version, with every update to it. If the next version arrives within a year of your purchase, your license covers it too; lifetime updates cover every version.

Scroll the page: what it says shows up here, the way captions do in the app.

This is the page talking. Nothing is recorded.