Speak Eazy

Privacy

Privacy policy

This policy says what personal data we hold, why we are allowed to hold it, who else sees it, and what you can ask us to do with it. It is worth reading, and it is written to be readable.

Last updated 21 November 2026

01Who we are, and what this covers

Speak Eazy is dictation that website owners install in their own pages, so that the people using those pages can speak into a text box instead of typing. It is operated by Speak Eazy, registered at Austin, TX 78750. For anything in this policy, write to privacy@speakeazy.dev.

This policy covers our website, the Speak Eazy console where our customers manage their account, and the dictation software our customers install. It does not cover the websites our customers run. Those are theirs, and their own privacy policy applies there.

For the data in sections 4 and 5, we are the controller: they are our own records of who holds an account and how much of the service was used. We are not a processor of anyone’s content, because no content reaches us. Section 3 explains how that works, since it is the unusual part and the part that matters most.

This policy is governed by the law of the State of Texas.

02Who “you” are, and who an end user is

Our software runs inside other companies’ websites, so two different kinds of people are involved. The difference between them decides nearly everything below, and we keep the two words apart everywhere in this document.

“You” means our customer

Someone who signs in to the Speak Eazy console. You have an email address with us, you administer one account, and you decide which of your own websites our software runs on. We hold personal data about you. Section 4 lists every piece of it.

An end user is somebody using your website

Someone who presses the microphone and speaks into a text box on a site that one of our customers runs. We hold no personal data about end users at all. We do not know who they are, we never give them an identifier, and we refuse one if it is sent to us.

So wherever this policy says “you”, it means our customer. Wherever it means the person dictating, it says end user, in full.

03What we never receive, and how you can check

Three claims follow. Each one is a consequence of how the software is built rather than a promise about how we behave, which is why we can describe the mechanism instead of asking to be trusted.

The audio

When an end user presses the microphone, their browser opens a connection straight to the company that turns speech into text, using a single-use key that our service issues for that one session and that stops working about fifteen minutes later. The sound travels from their browser to that company. It does not pass through us at any point.

You can see this for yourself: open the network panel in a browser on a page running our software and watch where the audio connection goes. The software is published openly on npm, so the code that opens it can be read line by line.

The words

The transcript is delivered to the same browser and inserted into the text box on your page. It is not sent to us, we do not store it, and we do not write it to a log. That includes the error paths, where this kind of thing usually leaks: our logging deliberately drops the fields that could carry it.

Anything that identifies an end user

No name, no email address, no identifier of yours, no device fingerprint. Our API validates every request against a fixed list of permitted fields, so a field we do not recognise causes the request to fail with a 400 error rather than being quietly saved. Nothing about an end user can reach our database by accident, because there is nowhere for it to land.

One scoped claim, since a vague one would be worse than none: we do store a hashed form of the network address a request arrives from, and your own code may optionally send a random per-browser value. Section 5 says exactly what both are, what they are for, and why neither identifies a person.

04What we hold about you

What For example Why we hold it
Your email address you@yourcompany.com To sign you in, and to reach you about your account. We do not add it to a marketing list unless you ask us to.
Your sign-in history Created 4 Jan, last signed in 20 Nov, address confirmed To let you back in, and to show you if someone else has been in your account.
Your account “Acme Ltd”, active, Starter plan To bill you correctly and to help you when you ask us to.
Your applications “Acme CRM”, key pk_live_…, allowed on acme.com To run the service, and to refuse your key if it turns up on a website that is not yours.
Your secret keys A one-way hash. Never the key itself. So your own server can start dictation. We cannot show you a secret key again after you create it, because we do not have it.
Your sign-in codes A one-way hash, valid for ten minutes To sign you in without a password. Used or expired codes are deleted nightly.

We ask for no postal address, no telephone number and no date of birth, because we would have no use for them.

05What we hold about each dictation

One row is written per dictation. These are the columns, named as they are in our database so that you can match this list against the real thing:

Column For example Why we hold it
application_id app_7Xk2… Which of your applications it belonged to, and so which account the words count against.
token_source publishableKey Whether the browser or your own server asked for it. Useful in support, and for spotting misuse.
origin https://acme.com The address of the page it ran on, so a key copied off your site cannot be spent somewhere else.
ip_hash A salted one-way hash. Never 203.0.113.7. To stop one source flooding the service. The address itself is never written down, and the hash cannot be turned back into it.
client_ref An optional random value your code may keep To tell one browser from another when applying abuse limits. It says “the same browser” and never “this person”. It is generated by your code, not ours, and you can leave it out entirely.
started_at, completed_at 20 Nov 2026, 14:03:11 To put the usage in the right billing month.
Word and second counts 32 words, 14 seconds This is what your bill is based on, and what you see in the console.

No column in that table identifies a person, and that is enforced rather than intended: a test in our codebase fails the build if a column for an end user is ever added to this table.

06Why we are allowed to hold it

Where the law requires a lawful basis for each use, as the UK and EU GDPR do, ours are these:

  • To perform our contract with you. Your email address, your account, your applications and the dictation rows. We cannot run the service, or bill you for it, without them.
  • Our legitimate interests. The hashed network address and the optional browser reference, both used to keep the service from being abused, and keeping dictation rows long enough to answer a billing dispute. We have weighed this against the people affected, and it is limited to data that identifies nobody.
  • A legal obligation. Invoices and the records behind them, which tax and company law require us to keep.
  • Your consent. Only if you opt in to hearing from us about the product, and you can withdraw it in one click without affecting your account.

The microphone itself is a separate matter: the browser asks the end user for permission, every time it is needed, and no recording can start until they grant it. That permission is between them and their browser. We are not party to it and we cannot bypass it.

07Cookies, and what sits in a browser

We set no cookies. Not on this website, not in the console, and not in the software our customers install. That is why there is no banner on this page asking you to accept any.

Two things do sit in a browser, and neither is sent anywhere else. When you sign in to the console, your sign-in is held in your browser’s session storage, which the browser discards when you close the tab. On a customer’s own site, our software stores nothing of its own and makes no network request at all until an end user presses the microphone. It sends us no analytics, ever, and there is no telemetry in it to switch off.

08Who else handles any of this

Who What they get Where
The company that turns speech into text The audio, directly from the end user’s browser, and the text it sends back to that same browser. Its own terms govern what it does with the audio. United States
The company that sends our email Your email address and the sign-in message we send to it. United States
The company that hosts the service Runs the service and the database holding sections 4 and 5. United States

We do not sell personal data, we do not share it for advertising, and we do not use it to train anything.

Because the audio reaches that transcription company rather than us, you should tell the people using your own website that their speech is transcribed by another company. We will name it, and everyone else we rely on, in our data processing agreement, so that you can check it and tell your own users. Our customer agreement sets out this division. If we add a company to the list, we will tell you 15 days before it starts.

09Where all of this is held

The service, its database and every company in section 8 are in the United States, so the United States is where all of this is held. If you are somewhere else, or the people using your website are, then using Speak Eazy means this data reaches the United States. We will say so here if that ever changes.

The audio is the case worth being precise about. It goes from the end user’s browser to the transcription company, in the United States, without passing through us. We make that connection possible by issuing the key for it, so we describe it here rather than leaving it out, even though we never hold the audio ourselves.

10How we keep it, and for how long

Everything travels over an encrypted connection. Secret keys and sign-in codes exist only as one-way hashes, so a copy of our database would not let anyone sign in as you or spend your quota. Keys for the transcription service are issued one at a time, work once, and expire in about fifteen minutes. An application only works on the web addresses you have allowed. And the largest part of the risk is simply absent: we cannot lose audio or transcripts, because we never have them.

What How long Then
Dictation rows 400 days Deleted automatically by a nightly job. That is a full year of billing plus the time a dispute takes to arrive.
Sign-in codes Ten minutes Deleted nightly once used or expired.
Your account and email address While your account is open Deleted 30 days after you close it, apart from invoices we are required to keep.

11What you can ask us for

You can ask us to show you what we hold about you, correct it, delete it, give you a copy to take elsewhere, or stop a particular use of it. You can also object to the uses that rest on our legitimate interests, and withdraw consent for product emails at any time.

Write to privacy@speakeazy.dev. We will answer within one month, and there is no charge. If you are not satisfied, and you are somewhere that has a data protection authority, you can complain to it.

One consequence of section 2 is worth saying plainly. Because we hold nothing about your end users, we cannot act on a request about one, and nobody can require us to hand over something we never had. If somebody using your site asks you what becomes of their speech, sections 3 and 8 are the answer, and you are the one who can give it.

12Decisions made by software

We do not profile anyone and we make no automated decision that affects a person. The only automatic decision in the service is whether an application has words left in its allowance, which is a fact about an account rather than about anybody’s character or circumstances.

13Children

Speak Eazy is sold to businesses. It is not meant for children, and we do not knowingly hold data about anyone under 5. If you think we have, tell us and we will delete it.

14When this policy changes

If we change what we collect, who else handles it, or how long we keep it, we will update this page and email the address on your account 15 days before the change takes effect. The date at the top says when this version was published, and a short note at the bottom says what moved.

15Contact us

privacy@speakeazy.dev, or by post to Speak Eazy, Austin, TX 78750. A person reads that address, and you will get a person’s reply.

Last updated 21 November 2026. First published version.