Terms of use
Voice Taker is a note taker for records that runs on your own device. These are the terms it is offered under — what is free, what is paid, what the software cannot promise, and the parts that are still to be settled.
- Last updated
- 1 August 2026
What Voice Taker is
Voice Taker listens to a record and writes it down on the machine in front of you. It separates the speakers, can translate as it goes, keeps the moments you mark, exports what it produced and reads a record back to you. On a paid plan it can also summarise a record you have switched sync on for. It builds for iPhone and iPad, Android, macOS and Windows.
It is not on the app stores yet. Nothing has been published to Google Play and no macOS or Windows build has been distributed, so today the only way to have it is a test build somebody sent you. These terms are written for the product as it will be offered rather than as it is distributed at this moment.
Every plain statement here was written from Voice Taker’s own source code and checked against it. Two kinds of paragraph are marked instead, because they are not that: a decision Voice Taker took, where the code decides nothing and somebody had to choose; and something only the owner of this product can supply, which is still missing and is marked as missing rather than guessed at.
Your account
You do not need an account to use Voice Taker. Everything the app does on your own device works without one. An account exists so that records you choose can be kept on the server, read on another device, and summarised by a hosted model.
Signing up asks for a name, an email address and a password, and nothing else. You are responsible for the password and for what is done through your account while it is signed in. If you think a device should no longer be signed in, the device list will sign it out from any other device you are signed in on.
An account is one person. Voice Taker does not sell a plan that several people share, and there is nothing in the product that adds a second person to an account — no place to invite anybody, and nothing an account holder can grant.
How old you have to be
You have to be sixteen or over to have a Voice Taker account. There is no minimum age for the app on your own device, because nothing it does there reaches Voice Taker or anybody else.
Signing up does not ask for a date of birth, so nothing in the product checks this. It is a rule about who may make an account, not a gate that stops one being made. The privacy policy sets out why sixteen and why only the account.
What is free and what is paid
Everything that happens on your device is free and needs no account: recording, transcription, telling the speakers apart, translation, bookmarks, export in five formats, and having a record read back to you.
Syncing the words of a record is free too. What a paid plan adds is two things: keeping the recording itself on your account, and summaries — which are written by a model too large to run on a phone, and are not made on the free plan at all.
On a free account the allowance for recordings is zero rather than small, and the product treats that as its own answer: an upload is refused with a message saying recordings are not part of this plan and that the recording stays on the device, rather than with an error about running out of room.
The prices, the storage allowances and the number of hosted summaries a month are rows in this server’s own catalog and are published with the plans. They are not repeated here on purpose: a number copied into a document is a number that drifts from the thing it describes without anybody noticing. The pricing section of this site and the Billing page in your account are where the current figures are, and those are the ones that apply.
Paying, renewing and cancelling
Payments are taken by Stripe. Card details are entered on Stripe’s own page and never reach Voice Taker’s server, which only ever asks Stripe for a checkout page and for a link to Stripe’s customer portal.
A subscription renews until you cancel it. You cancel through the customer portal, which the Billing page in your account links to. A subscription set to end at the period end keeps working until then rather than stopping immediately, and a payment that has failed keeps access until the period it was for runs out.
Your plan changes only when Stripe tells this server it has. No request the app or this website can make grants an account anything.
Refunds
Cancelling stops the next payment and leaves you with what you have already paid for: access continues to the end of that period. There is no refund of a period that has already begun, and no partial refund of an unused part of one.
If the law where you live gives you a right to cancel and get money back that these terms do not, that right applies and nothing here removes it. A refund made for any reason is made by hand through Stripe by whoever runs Voice Taker; there is no button in the product that issues one.
If you are paying, cancel before you delete your account. Deleting the account removes the link the console uses to reach the payment page, and it does not tell Stripe to stop. The delete screen says the same thing before it lets you go ahead.
Recording other people
A record has other people in it. Voice Taker records everyone in the room, not only the lecturer, works out which utterances belong to the same voice, and lets you give that voice a name. It keeps a voiceprint of each speaker on the device so that it can tell the same people apart from one record to the next — the privacy policy sets out what that is, where it stays and what removes it.
Whether you need those people’s permission, and whether you have to tell them, depends on where you are and on the rules where you study — many universities have their own policy on recording lectures. Voice Taker cannot know either. It does not ask you where you are, it does not check, and it will record whatever the microphone can hear.
And signing in is a second decision
Everything above stays on your device unless you sign in. Signing in on a device puts the records on it — other people’s words, and any name you typed for them — onto Voice Taker’s server, and from then on it happens without a further tap. That is still your act rather than something the app decided, you can stop it from the app’s settings or by signing out, and you are responsible for it in the same way.
There are two steps further, and each is its own decision. 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. Asking the assistant about your records sends the sentences it finds to the same company.
If somebody in a record asks you to get rid of what was recorded of them, what exists is on your side rather than theirs: they have no account here and nothing to ask Voice Taker for. Deleting the recording, or deleting the record, takes their words, their voiceprint and any name you gave them with it — on the device and on the account.
How Voice Taker may be used
Voice Taker is for writing up records you are entitled to record. Using it for anything the law where you are forbids is not allowed, and neither is any of the following.
- Recording people where the rules that apply to you say you may not — see the section above, which is where that responsibility sits.
- Putting anything on your account that you have no right to hold, or that it is unlawful to hold.
- Trying to reach another account’s records, recordings, summaries or billing, or trying to get administrator access.
- Working around the request limits, running Voice Taker’s server from a script, or using an account to move volume rather than to keep your own records.
- Reselling Voice Taker, or offering it to other people as a service of your own.
- Attacking the service or the people using it — probing it for holes, interfering with it, or sending anything designed to damage a device.
Two of those are enforced by the software rather than only asked for here. Every route on this server has a request limit, sized against what a real device does, and a route with no limit written down stops the server starting. And every read is scoped to the account that asked for it, so reaching another account’s records is refused by the server rather than merely discouraged.
That is the whole of the mechanism, and it is worth knowing how small it is: there is no suspension feature in this product and no screen that sets one. The one lever is a single column in the database, set by hand, which the sign-in check reads.
Ending it
You can stop using Voice Taker at any time. Turning sync off in the app’s settings, or signing that device out, stops anything further being sent; deleting a record removes it from the account; and deleting your account removes the account and everything reachable from it, immediately and permanently, from inside the app or from this website.
One record survives a deletion — the payment events as the payment company sent them, no longer linked to you. The privacy policy says exactly what that is and why. And one case is refused rather than performed: an account that has used the administrator console has written rows in an audit log that must not lose the name of who acted, so that deletion has to be settled by hand first.
Nothing on your own device depends on any of this. A record you never synced is unaffected by anything Voice Taker does to an account, because it was never anywhere else.
What the software cannot promise about accuracy
A transcript is a machine transcription. It mishears words, and it mishears them more in a noisy room, with an unfamiliar accent, or when several people talk over each other. Speaker labels are a guess from the sound of a voice and can put two people together or split one in two.
Translation is worse than that, and this is measured rather than a caution. The model on the device does not simply make mistakes — it invents, fluently, and it gets numbers wrong: a page or a chapter can come out as a different one and read just as plainly as the rest. The app says so on the screen where the choice is made, before it is made.
Summaries leave things out. Measured against one ninety-minute record with six facts a lecturer stated plainly, the hosted model recovered all six on every one of eight runs — which means the summaries are worth reading, and does not mean any one of them is correct. A summary is shorter than the record it is made from, and the transcript is what was said.
So: do not rely on a transcript, a translation or a summary for anything that matters without checking it. Voice Taker gives you the timings and, if you kept it, the recording, precisely so that checking is possible.
Availability, and keeping your own copy
No promise is made anywhere in this product about how much of the time the service is reachable, and none is made here. Nothing you do on your own device depends on it: the app records, transcribes, searches and exports with the connection off.
Two things about recordings on your account
There is no way to download one back. The server can be given a recording and can say how much of it has arrived; it has no route that returns the audio, and no part of the app or this website asks for any. A recording on your account protects you against losing the device and does nothing else.
And the recordings are not backed up. The database is dumped nightly and kept for up to a fortnight, so the words of a record are; the audio is not, because a second copy of the space set aside for recordings does not fit on the machine. If a recording held on the account is lost, the copy on your device is the only one.
So keep your own copy of anything that matters. The app exports a record as plain text, subtitles, a Word document, a PDF or the recording itself, and that is the way to hold something that does not depend on this service or on this device.
What is guaranteed, and what is not
Voice Taker is offered as it is. There is no promise that it will be available when you want it, that it will run without faults, that it will work on a particular device, or that what it produces is accurate — the section above says in measured terms how inaccurate it can be.
What Voice Taker is not responsible for
- A record it failed to record, or recorded only part of.
- A transcript, a translation, a speaker name or a summary that is wrong — and anything you decided on the strength of one.
- A recording or a transcript lost from your device, or lost from your account. The recordings are not backed up and cannot be downloaded back; both are said above, and both are why exporting a record you care about is the answer.
- What a company that wrote a hosted summary does with the record you asked it to summarise. Their terms apply to their copy.
- Anything that follows indirectly rather than directly — a missed deadline, an exam, an opportunity, a lost profit.
What is not excluded
Nothing here limits responsibility for death or personal injury caused by negligence, for fraud, or for anything else the law where you are does not allow to be limited. If you are a consumer, you keep every right the law gives you and nothing in this document takes one away.
This is the one section of these terms not written from Voice Taker’s own source code, because there is no source code for it: it is a standard clause, and whether any part of it holds depends on where you are, whether you are a consumer and what the law where Voice Taker is established allows — which is one of the questions still unanswered at the foot of this page. Expect this section to be rewritten before it is relied on.
The parts that are not written yet
Three questions below are not statements about software and cannot be answered by reading it: who exactly you are agreeing with, whose law decides a disagreement, and where to write. None of them is recorded anywhere in this product, and each is left blank rather than filled with something plausible — a paragraph nobody chose is worse than a marked gap, because nobody ever goes back and reads it again.
Changes to these terms
These terms are written from the product’s own source and are meant to change when the product does. The date at the top is the last time they were checked against the code, not the last time a word was edited.