Complete
Every word, not your partial notes. A fast live pass appears as the lecturer speaks, then a slower and more accurate one settles behind it.
Your notes, taken for you
Rest your mind, leave it to Bird.
Press record, put your pen down. Every word is written as it happens and the voices are told apart — all on the device in your hand.
- AI features
- Speaker detection
- Plug and play
Available for
- iPhone
- Android
- Mac
- Windows
How a record becomes notes
Four stages between a lecturer speaking and a file you can hand in.
- Record
- Transcribe
- Speakers
- Export
On device
All four stages run with the network off.
What you are promised
Three things at once.
Cloud transcription usually makes you trade one of these for another.
Private
Speech recognition, speaker separation, translation and reading a record aloud all execute locally. There is an upload step, and signing in is what starts it — your records then sync to your account, and one switch in the app’s settings turns that off. Nothing is uploaded unless you sign in. Making a summary of a synced record sends its words to the company whose model writes it.
Offline
The models live on your phone, so a basement lecture hall with no signal behaves exactly like a seminar room with full bars.
How it works
Every step runs on the device in your hand.
Audio goes in and a written record comes out.
All of this, on your device
- The room
Press record once, then put the phone down.
- Live text
Words land as they are spoken, then a slower pass corrects them.
- Speakers
Every voice gets its own labelled thread. Nobody tags anything.
- Your library
Every word of every record, searchable with the network off.
The app
Six screens, one job.
The record happens once. Everything here exists to hand all of it back to you in writing.
LibraryEvery record you have kept
Search runs against the database on the device, so a result appears with the radio off and with nothing to wait for.
- Folders, favourites, and a bookmark on any moment of any record.
- Recently Deleted holds anything you remove for thirty days.
- A fresh install is genuinely empty. There is no sample lecture in it.
Live capturePut the phone down and listen
Text lands while the lecturer is still talking. Capture is text only unless you say otherwise, and keeping the audio is a choice you confirm for that record before it starts.
- Pause, resume, and choose which microphone before you begin.
- One tap across the control bar marks the moment, and it works while paused.
- A record interrupted by a crash or a flat battery comes back.
TranscriptWho said what, without tagging anyone
Voices are separated as the recording runs, and each one gets its own labelled run. Name somebody once and Voice Taker finds them again in the records after it.
- Your corrections stay marked as yours, separate from what the machine wrote.
- A record in another language can be written down in English — chosen before it starts, with the machine named as the author.
- Export as plain text, subtitles, Word, PDF, or the recording itself.
SummariesWhat it is, and what it is not
A summary is part of a paid plan. It is written by a model too large to run on a phone, so it reads the copy of the record on your account — which means it can only see a record that has synced there. The app asks before it sends one, every time.
- Nothing is summarised unless you ask for that record, and the app says so first.
- A summary sits beside the transcript and never replaces it.
- A summary is shorter than the record, so it leaves things out. The transcript is what was said, and it stays.
SettingsWhat is on this device, and what it costs
Running the whole thing on a phone means the models live there too. This is where they are listed, with what each one takes up and what removing it would cost you.
- The first run downloads 317 MB: finding speech, writing it down, and telling the people in the room apart.
- Reading a record aloud is a separate download, and nothing installs it for you.
- Sync lives here too: on from the moment you sign in, for every record, and one switch here turns it off.
First runThe download, said out loud first
Voice Taker cannot write a word down until its models are on the device. The first screen says so and names the size, rather than letting you find out in a lecture.
- 317 MB, quoted before the first byte, and the same 317 MB for everybody.
- The microphone is explained before it is asked for. The system prompt is never the first thing you see.
- Nothing here asks who you are. Recording, transcription and search never do.
Everything it does
28 featuresThe whole product, not the highlights.
Pick a category to see what each one does.
- Live transcriptionText appears while the lecturer is still talking.
- Two pass accuracyA fast live pass, then a slower confirmed one behind it.
- Text only by defaultKeeping the audio is a per record choice, confirmed first.
- Pick the microphoneChoose your input before the record begins.
- Pause and resumeStep out of the room without ending the record.
- Survives a crashAn interrupted record comes back, even after a dead battery.
- Eight live languagesArabic, English, Indonesian, Japanese, Russian, Thai, Vietnamese, Chinese.
- A wider offline setFinal transcription across many more, or any BCP-47 tag by hand.
- Right to left, typesetArabic is set properly rather than mirrored, everywhere it is written out.
- Translate as you recordEnglish from a record in another language, with the machine named as the author.
- Automatic separationEvery voice gets its own labelled thread. Nobody tags anything.
- Rename anyoneFix the labels after the record, once you know who spoke.
- Known voicesName somebody once and they are found again in the records after it.
- Instant offline searchEvery word of every record, with the network off.
- BookmarksMark a moment with one tap while the record is running.
- Folders and favouritesKeep a whole semester in an order that makes sense.
- 30 day recoveryRecently Deleted holds anything you removed by mistake.
- Tracked editsYour corrections stay separate from the machine output.
- Read aloudHave a record read back to you, on your device.
- Export in five waysPlain text, subtitles, Word, PDF, or the recording itself.
- SummariesWritten by a model too large to run on a phone, on a paid plan.
- The paid one needs a synced recordIt reads the copy on your account, so it can only see a record that has synced there. Making a summary of a synced record sends its words to the company whose model writes it.
- It says what wrote itA summary carries the model that produced it, and what it is likely to leave out.
- Sync, from the moment you sign inOn once you sign in. Your records reach your other devices on their own, and one switch in the app’s settings turns it off.
- Recordings on your accountThe paid line: the audio itself kept off the phone, for the records on your account.
- Every device you sign in onSee them, and sign any one of them out from any other.
- Read a record in a browserThe records that have synced, with their speakers and their timings.
- Delete the whole accountFrom inside the app or the browser, with a receipt that names nobody.
Apple — iPhone, iPad and Mac
Android
Windows
Phones and tablets
Laptops and desktops
Your account, in a browser
Runs on every device you own
One app for iPhone, iPad, Mac, Android and Windows — and once you sign in, your records open in a browser too.
See the screensWhat runs on your device
Pricing
Choose your plan.
Get Pro features to maximise your experience.
Reading the price list…
Questions
The four people ask first.
Does it upload my records?
Not until you sign in. Capture is text only by default, and keeping the audio is a per record choice. Nothing reaches a server unless you sign in — once you do, your records sync to your account, and one switch in the app’s settings turns that off. Voice Taker is not always the last stop: making a summary of a synced record sends its words to the company whose model writes it.
Do I need an account?
Not to record, transcribe, search or export — none of that asks who you are. An account is what sync needs: the records on a signed-in device reach your other devices, and nothing else does.
How good is any of this?
Good enough to read back, and not good enough to trust blind. The live pass is corrected by a slower one behind it; a translated record says a machine wrote it; and a summary is shorter than the record it is made from, so the app says where the transcript is rather than leaving you to find out.
What does it cost?
Nothing, for everything that happens on your device. What a plan buys is room on your account for the recordings themselves, and summaries written by a model too big to run on a phone.
Your next lecture happens once.
Free to start, no account, and not one byte of it leaves the device in your pocket unless you sign in.