AudioClerkUpload a recording

A product of A.I. Tech Inc.

Someone to keep the record.

Send us a recording and get back a transcript with every speaker named and every turn timestamped — by a time you pick. No account, no subscription, no password. Your audio is destroyed 24 hours later, and we keep the log that proves it.

From $3 a recording. No subscription, no minute bundles, nothing to cancel.

Who said what

Speakers separated automatically, then named by you in a few seconds.

By a time you choose

You buy a finished document by a deadline, not an amount of processing.

Deleted in 24 hours

Not a policy — the machine that runs your audio has no network at all.

Six formats out

Markdown, plain text, SRT, VTT, TSV and word-level JSON with confidence.

How it works

Four steps, and you are only present for the first one.

  1. 01

    Drop the file

    Fourteen audio and video formats, nothing to convert. We read the length and quote you before you pay.

  2. 02

    Pick what you need

    Standard comes back in minutes. Overnight costs less because you let us run it in the quiet hours.

  3. 03

    Close the tab

    Your card is authorised, not charged. We email you a private link and tell you when it is ready.

  4. 04

    Name the speakers

    Type real names over SPEAKER_00 and the whole document re-renders. Then download it, six ways.

What it costs

Priced per recording, not by the minute. You are buying one finished document, so that is what the invoice says — and there is no subscription to forget about.

Quick

$3

per recording

A recording up to 15 minutes, back in minutes

For a voice note, a short call, a single interview segment.

Our own GPU

Standard

$7

per recording

Up to 4 hours, back within the hour — usually minutes

Speakers separated and named, six export formats, no subscription.

Our own GPU

Overnight

$4

per recording

Up to 4 hours, back by 9am tomorrow

The same transcript for less, because you let us run it in the quiet hours. Nothing is held back — you are simply not paying for a slot right now.

Our own GPU, when it would otherwise be idle

Add to any of them

Add-ons attach to whichever package you picked — including Overnight, so private and unhurried is a real combination rather than a choice between two things you need.

Attested

from +$19

Proof you can hand to someone else

  • up to 30 min+$19
  • up to 2 hours+$29
  • up to 4 hours+$49

Every recording already runs on hardware we own, in a container with no network. Attested is what you add when you have to demonstrate that to an auditor, a regulator, a client or an insurer: a signed DPA, the deletion-log entry delivered with your transcript, an access log of who opened the link and when, a passcode on by default, and upload routed straight to our worker so it never rests on a rented machine.

Briefed

+$12

Summary, decisions, and action items with owners

The thing you actually wanted from the meeting. Priced to be an easy yes rather than a considered purchase — it costs us cents and it is the only part of this nobody has commoditised.

Verified

+$2.00/min

A person reads it against the audio

Verified, not certified. We are not court reporters and this is not a sworn record — but a human has checked every line.

Ten recordings or more drops to $3 each, fifty or more to $2. For hundreds of hours — a council's minute book, an oral-history collection, a decade of board meetings — talk to us about a project price.

Why we can promise deletion when others cannot

Most transcription tools send your audio to somebody else's API. We run the whole pipeline on GPUs we own, in a container with no network connection at all. The models are baked into the machine image before it starts, so the container has nothing to phone home to.

That is not a policy we could quietly change. It is a property of how the thing runs — your audio physically cannot leave. And it applies to every package: there is no upgrade that makes your recording more private, because there is no version of this that sends it anywhere.

Read the retention table

Your audio: 24 hours

Destroyed a day after the job finishes. Change it to immediate, 7 days or 30 if you prefer.

No vendor in the path

Every package runs on hardware we own. There is no transcription API behind us, at any price.

And provable, if you need it

Add Attested for the signed DPA, the deletion-log entry, and an access log — for when you have to show someone else.

Decoded audio: never written

Working files live in memory and die with the container. Not configurable, because it should not be.

No voiceprints, ever

Speaker embeddings are never saved and never matched across recordings.

Excluded from backups

Deliberately. A lost disk means lost recordings — the right trade for this product, and we say so.

What we will not do

Worth knowing before you upload, rather than after.

  • We are not court reporters.

    Machine transcription is roughly 85–95% accurate on real conversation with crosstalk and accents. Our checked tier is called verified, not certified, and there is no chain of custody.

  • We do not take health data.

    No BAA is offered. Clinical recordings are out of scope — plainly, and in writing, rather than discovered later.

  • We never build a voiceprint.

    Matching a voice across recordings is biometric data under GDPR and Illinois BIPA. Speakers are matched within one recording and nowhere else.

  • We do not say unlimited.

    We publish the real number of hours. A business with a fixed number of machines that promises unlimited is choosing who to disappoint later.

  • We will not knowingly process covert recordings.

    You attest at upload that you had the right to record. We will send you a recording-notice script to read at the top of a call.

Try it with one recording.

Seven dollars, back before you have finished your coffee, and if we cannot transcribe it you are not charged.

Upload a recording