The real cost of cloud transcription over three years

You are choosing between paying every month and paying once. The monthly number looks small and the one-time number looks large, which is exactly the wrong way round after eighteen months.

Here is the arithmetic written out, including the part of the local bill that no pricing page mentions.

All figures come from our own pricing pass in July 2026, taken from official pricing pages and second-source roundups. Anything we could not verify is left out rather than estimated.

The two shapes

Product	Price	Shape
Otter / Fireflies / Krisp cluster	~$100–200/yr	Subscription
Meetily Pro	$120/user/yr, 1 device	Annual license
Plaud Pro	$100–216/yr (+$159 hardware)	Subscription + device
Plaud Unlimited	$240–360/yr (+$159 hardware)	Subscription + device
VoiceInk	$25–49	One-time
MacWhisper Pro	€59	One-time
Audio Hijack 4	$69	One-time, paid major upgrades
yapyap	$69	One-time, 2 devices
Voibe	$149	One-time
Superwhisper Pro Lifetime	$249.99	One-time

Two things fall out of that table.

The cheapest subscription costs about the same in year one as the most expensive one-time application. By year three it has cost four to five times as much, and you own nothing at the end of it. The Plaud rows are worse than they look, because the subscription sits on top of $159 of hardware you already bought.

And Superwhisper's $249.99 "lifetime" tier is priced against its own $8.49 monthly plan, so it is a subscription buyout rather than a comparable one-time price.

The part nobody invoices you for

Running the model yourself moves costs off the invoice and onto you. Three of them.

Electricity. A local transcription run pins your processor for the duration and that draw is real. We have not measured it on a representative machine, so we are not going to quote a figure. Anybody who quotes one without a benchmark behind it is guessing.

Your time and your laptop. Work that a cloud service does on a rack somewhere happens on your machine instead, competing with everything else you are running. A long recording finishes while you wait rather than while you walk to lunch.

Disk. Speech models and local language models are gigabytes each. Keep several for different jobs and it adds up quickly.

Underneath all three sits the machine itself. Local transcription assumes you own hardware fast enough to run these models at a speed you can live with. If you do not, the one-time price is not the whole price, and a subscription may genuinely be cheaper for you.

What the subscription actually buys

It buys somebody else's hardware doing the work while your laptop is shut. It buys a two-hour recording finishing without your machine getting hot. It buys models improving without you downloading anything, and never thinking about disk space.

Those are real goods. Pretending otherwise would make this post useless to you.

What it does not buy is a copy you keep. Stop paying and the archive stops being reachable. That is the trade, and it is worth naming in the same breath as the price.

Where the arithmetic points

MacWhisper Pro at €59 is cheaper than yapyap, established, well liked, and does local transcription with speaker labels and meeting capture, with a free tier underneath. If that is the whole of what you need on a Mac, buy it. The arithmetic points there and it should.

yapyap is $69 once for two devices, which is less than a single year of Meetily, Plaud or the Otter cluster. The ten dollars over MacWhisper is meant to pay for three things: your phone acting as the microphone and handing the file to your desktop over your own network, speaker labelling that keeps the call and the room on separate tracks, and write-ups you can author and share with other people.

Against the subscription rows in that table, either one pays for itself inside the first year. The difference is that neither of them asks again in year two, and the top of the table does, forever.