Changelog
Every release of Transcriber for macOS, newest first. The app checks for these once a day and can install them itself.
Transcriber 0.6.3
⚠️ Coming from 0.5.0? Please download the new version once by hand
This one time, Install and Relaunch will stop with "Update refused". Nothing is wrong with your copy, and nothing is wrong with this update.
Why: Transcriber is a side project of mine, still early, and I'm not a registered Apple developer yet. So every version is signed with a homemade certificate instead of one from Apple. Your installed copy only accepts updates that carry the same certificate. That's a safety check, so nobody can slip you a fake update. I lost that original certificate when my Mac's keychain was reset, so from 0.6.0 on the app carries a new one, and your copy rightly refuses it. Sorry for the extra step. I've put safeguards in place so this doesn't happen again.
What to do (two minutes):
- Download the Transcriber .zip from the release page and unzip it.
- Quit Transcriber (menu bar icon → Quit), then drag the new Transcriber.app into your Applications folder and replace the old one.
- Open it. If macOS blocks it as coming from an unidentified developer, go to System Settings › Privacy & Security and click Open Anyway.
- macOS asks once more for microphone and system audio access. Your recordings, settings and remembered voices stay where they are.
- If you use auto-paste into your AI tool: open System Settings › Privacy & Security › Accessibility, remove Transcriber with – and add it again with +. The old switch looks on, but it still belongs to the old version, so macOS would otherwise keep asking.
After that, updates install with one click again.
What's new in 0.6.3
- Fixed calls that recorded nothing from the far end when Transcriber started together with the call (FaceTime): the call's audio was never picked up, and the waveform stayed flat. System audio is now watched: when something plays but nothing arrives, the capture is rebuilt on its own within seconds.
- New "Reconnect Audio" in the menu bar menu while recording: restarts microphone and system audio without stopping the recording.
- The far-end voices are told apart live: the transcript card shows "Speaker 1", "Speaker 2", … in their own colours, also when two people talk in one breath, and a remembered voice is named right away.
- Naming is per voice now: the card's bottom bar shows who is talking, type the name next to it and Enter. Every line of that voice gets the name, earlier ones too, and no one else does. Click a name in the transcript to name that voice instead.
Also new since 0.5.0
- 0.6.2: when auto-paste into your AI tool fails for lack of the Accessibility permission, the dialog explains why and opens the right settings.
- 0.6.1: Recording flags given at launch now ask "Start a recording?" first, and auto-paste only goes into your AI tool while it is in front.
- 0.6.0: smaller floating pill with mute and pause, live transcript on hover, naming speakers, remembered voices, a Recordings section, automatic audio clean-up and more. The full list is under Settings › General › Show Changes.
Transcriber 0.6.2
⚠️ Coming from 0.5.0? Please download the new version once by hand
This one time, Install and Relaunch will stop with "Update refused". Nothing is wrong with your copy, and nothing is wrong with this update.
Why: Transcriber is a side project of mine, still early, and I'm not a registered Apple developer yet. So every version is signed with a homemade certificate instead of one from Apple. Your installed copy only accepts updates that carry the same certificate. That's a safety check, so nobody can slip you a fake update. I lost that original certificate when my Mac's keychain was reset, so from 0.6.0 on the app carries a new one, and your copy rightly refuses it. Sorry for the extra step. I've put safeguards in place so this doesn't happen again.
What to do (two minutes):
- Download the Transcriber .zip from the release page and unzip it.
- Quit Transcriber (menu bar icon → Quit), then drag the new Transcriber.app into your Applications folder and replace the old one.
- Open it. If macOS blocks it as coming from an unidentified developer, go to System Settings › Privacy & Security and click Open Anyway.
- macOS asks once more for microphone and system audio access. Your recordings, settings and remembered voices stay where they are.
- If you use auto-paste into your AI tool: open System Settings › Privacy & Security › Accessibility, remove Transcriber with – and add it again with +. The old switch looks on, but it still belongs to the old version, so macOS would otherwise keep asking.
After that, updates install with one click again.
What's new in 0.6.2
- When auto-paste into your AI tool fails because the Accessibility permission is missing, the dialog now explains why (a switch that looks on can still belong to an older version) and opens the right settings with one click.
Also new since 0.5.0
- 0.6.1: Recording flags given at launch now ask "Start a recording?" first, and auto-paste only goes into your AI tool while it is in front.
- 0.6.0: smaller floating pill with mute and pause, live transcript on hover, naming speakers, remembered voices, a Recordings section, automatic audio clean-up and more. The full list is under Settings › General › Show Changes.
Transcriber 0.6.1
⚠️ Coming from 0.5.0? Please download this version once by hand
This one time, Install and Relaunch will stop with "Update refused". Nothing is wrong with your copy, and nothing is wrong with this update.
Why: Transcriber is a side project of mine, still early, and I'm not a registered Apple developer yet. So every version is signed with a homemade certificate instead of one from Apple. Your installed copy only accepts updates that carry the same certificate. That's a safety check, so nobody can slip you a fake update. I lost that original certificate when my Mac's keychain was reset, so from 0.6.0 on the app carries a new one, and your copy rightly refuses it. Sorry for the extra step. I've put safeguards in place so this doesn't happen again.
What to do (two minutes):
- Download Transcriber-0.6.1.zip from the release page and unzip it.
- Quit Transcriber (menu bar icon → Quit), then drag the new Transcriber.app into your Applications folder and replace the old one.
- Open it. If macOS blocks it as coming from an unidentified developer, go to System Settings › Privacy & Security and click Open Anyway.
- macOS asks once more for microphone and system audio access. Your recordings, settings and remembered voices stay where they are.
- If you use auto-paste into your AI tool: open System Settings › Privacy & Security › Accessibility, remove Transcriber with – and add it again with +. The old switch looks on, but it still belongs to the old version, so macOS would otherwise keep asking.
After that, updates install with one click again.
What's new in 0.6.1
- Recording flags given at launch (--record-test, --capture-test, --start, --cycle) now ask "Start a recording?" first, so no other program can record through Transcriber's microphone permission.
- Auto-paste into your AI tool only happens while that app or browser is still in front; if you switched away, the text stays on the clipboard instead of landing in the wrong window.
Transcriber 0.6.0
- The floating pill is back on by default and much smaller: a waveform shows that audio arrives, with a mute button on the left and pause on the right. The menu bar icon simply turns red while recording; clicking it stops.
- Mute your own microphone from the pill (or the menu) to talk to someone in the room: only the computer's audio is recorded and transcribed until you unmute.
- Hover the pill to read the live transcript: a scrollable card unfolds below it with the whole conversation so far. The pill and its buttons stay put. Click the waveform to keep the card open.
- More accurate live text: speech is recognised in whole sentences with a few seconds of context, and the newest words stay blurred (sometimes several lines) until the model is sure of them.
- Name whoever is speaking: type a name at the bottom of the transcript card (best for longer turns like a talk). After the call that voice gets the name across the whole recording, also where it was not named. When the card opens on hover the cursor is already in the name field, so you can type straight away. Replaces the "Speaker 1 / 2" naming pills (live and in the save prompt) and the experimental live speaker separation.
- Meetings with many participants are now split into speakers properly. Before, anyone with less than 15 % of the talk time was merged away, so a 20-person call came out as a single "Speaker 1".
- Fixed system audio recorded at double speed and half length on AirPods (and other devices that change their sample rate mid-call), which also ruined speaker separation and echo removal for those calls.
- Transcriber remembers voices: once you name a speaker in the pill, they are recognised and named automatically in later recordings. Remembered voices (voice prints only, no audio) can be forgotten one by one or all at once in Settings → Transcription.
- New Recordings section (Settings, or "Recordings…" in the menu): every past recording with its audio status. Re-transcribe a recording with the current settings, or drop an audio file to transcribe it as a new recording.
- Raw audio is now kept for 7 days by default and then deleted automatically (or choose: delete when saved, 30 days, until you delete it). The transcript always stays.
- Phases with your microphone muted are marked in the transcript.
- Your side and the other side stay in sync: when one channel briefly delivers no audio (system audio starting late, a device switch), the gap is filled with silence instead of shifting everything after it.
- More languages to choose from (Settings → Transcription); the setting now explains what it can and cannot do.
- Echo removal works during the call again: without headphones, the far end coming out of your speakers no longer shows up as your own (blurred) text in the live transcript. The gate never locked on before, because it aligned the two channels by a clock that could be off by more than it searched.
- Your microphone now moves the waveform as much as the computer's audio does; before it barely moved unless you spoke very loudly.
- The live transcript card opens right at the latest line instead of visibly scrolling down from the top, and the "Latest" button appears whenever you are not at the end and is easy to read over the text.
Transcriber 0.5.0
Name speakers from the floating bar. Hover the bar during a call and pills appear beneath it, one per remote speaker; type a name from the keyboard (regardless of the frontmost app), Enter confirms, Tab moves between them. Names apply for that recording, in the bar and the saved transcript, and the same pills appear under the "Name this recording" prompt at the end. The naming pill is available from the moment a recording starts, and a name you type is kept even if you press Stop straight away. Experimental live speaker separation (Settings → Speakers): tells remote speakers apart during the call and labels them Speaker 1, 2, … as they appear. Less accurate than the after-call pass and downloads a model on first use; when on, it replaces the after-call pass. The remote speaker count is now detected automatically (no more "Max remote speakers" setting). Copy/paste (⌘C/⌘V, ⌘X/⌘A/⌘Z) now work in the Settings and naming text fields. Auto-paste into the AI tool no longer asks for Accessibility permission on every paste.
Transcriber 0.4.0
Re-transcribe after the call: re-runs the speech model over the recorded audio in long context windows and replaces the live transcript. Noticeably more accurate on quick back-and-forth speech. Enable it in Settings → Transcription (off by default). Microphone no longer goes silent after a mid-call audio-device switch: the input engine is restarted when it stops delivering audio, not only when it delivers silence. Better microphone level on multi-channel input devices: when the voice sits on one channel, that channel is used instead of an average that buried it.
Transcriber 0.3.10
Floating bar now appears every time you start a recording with the setting enabled, including when the first recording after launch has to load the speech model.
Transcriber 0.3.9
Floating bar: no more faint rectangle behind the paused pill; the capsule shadow stays
Transcriber 0.3.8
Fixed the recording start failure (Core Audio error -10868) after Apple voice processing falls back to plain capture. The fix announced in 0.3.3 had not made it into the build.
Transcriber 0.3.7
The close button of the Settings and What's New windows sits on the left, like the traffic lights.
Transcriber 0.3.6
System audio capture now works when the output device is a USB audio interface with its own inputs (e.g. Focusrite Scarlett): the tap's stream is read instead of the interface's silent input channels.
Transcriber 0.3.5
New What's New window (Settings → General → Show Changes) listing every version and its changes.
Transcriber 0.3.4
Settings → General gains an About section with links to the GitHub repository, issue reporting and GitHub Sponsors.
Transcriber 0.3.3
Fixes a recording start failure (coreaudio error -10868) after Apple voice processing falls back to plain capture, and removes the dark rectangle that could linger behind the floating bar after the save prompt.
Transcriber 0.3.2
Copy Prompt no longer resizes the template field and shows a short “Prompt copied” pill instead.
Transcriber 0.3.1
Adds a Copy Prompt button below the prompt template in Settings → AI.
Transcriber 0.3.0
Redesigned Settings window with light and dark mode (new Theme setting under General). The recording title is now entered directly in the floating bar, which morphs into a save prompt after stopping; all bar buttons are pill-shaped.
Transcriber 0.2.4
Security hardening release.
- Verified updates: the built-in updater now checks that a downloaded update is signed by the same identity as the installed app before installing it, and no longer removes the quarantine flag. Ad-hoc signed builds cannot self-update; sign with a stable identity (see README).
- Private clipboard hand-off: when a transcript is handed to your AI tool, it is marked as concealed/transient on the clipboard (skipped by clipboard managers and Universal Clipboard) and cleared shortly after the automatic paste.
- Hardened runtime: the app is signed with the hardened runtime and the unused library-validation bypass entitlement is gone.
- Logs no longer contain transcript fragments or guessed speaker names.
- Relaunch after an update no longer goes through a shell.
No changes to recording, transcription or output formats.
Transcriber 0.2.3
On first launch Transcriber now offers to download the speech models right away instead of at the start of the first recording. The automatic update check interval is configurable in Settings: never, every launch, daily, weekly or monthly.
Transcriber 0.2.2
New update window in the style of the Settings window: app icon, installed and available version, full release notes and an Install and Relaunch button, replacing the small popup. Checking, up-to-date and error states use the same window.
Transcriber 0.2.1
Fixes: the Name-this-recording dialog could get stuck when a recording was started while the previous one was still being saved. The floating bar no longer shows a faint rectangular backing on some Macs. The menu bar icon is white on a dark menu bar. Quitting during a recording no longer hangs for 15 seconds.
Transcriber 0.2.0
Transcriber 0.2.0
The app is now called Transcriber (repository renamed from AnyRecord; existing installs keep updating).
Live transcript bar
- Calm, stable text: words never jump lines or repeat, speaker names only when the speaker really changes
- Echo of the far end no longer shows up under your name; wrongly shown words are taken back
- The bar morphs: compact pill while loading / paused, full width with text, and it turns into the "Stop recording?" confirmation when you click the menu bar icon
Faster start
- Speech and VAD models are loaded once at launch and kept in memory – recording starts immediately instead of after "Loading models…"
Menu bar
- New glyph after the app icon; pulses red while recording
- Smoother expand / collapse animation
- Stop confirmation instead of stopping on a stray click
AI hand-off
- transcript.md is attached as a file, the prompt is pasted as text
- New default prompt: meeting type, summary, to-dos by person and urgency, open questions
Other
- "Open Last Transcript" reveals the file in Finder and opens Quick Look
- Experimental: guess remote speakers' names from the conversation ("[Anna]")
- Redesigned Settings; options for Dock icon and app-wide sound effects
- Title dialog focuses its text field
AnyRecord 0.1.1
First tagged release: local call transcription with echo gate, pause button and in-app updates.