What an hour of transcription actually costs, compared

An hour of machine transcription costs between roughly €0.27 and €12, and the spread is almost entirely about whether you use the allowance you are paying for. Below is every plan turned into a cost per hour, with the break-even points, so you can find your own volume on the table instead of guessing.

All prices checked on 23 September 2026 and quoted in the currency each vendor charges. They change — check before you buy.

The plans, as cost per hour

Each plan's monthly price divided by the hours it includes, assuming you use the allowance.

Two things jump out of that table, and neither is on any pricing page.

Why does the expensive Descript plan cost more per hour?

Because you are not buying hours. Creator at $24 includes 30 media hours — about $0.80 an hour. Business at $50 includes 40 — about $1.25. Paying twice as much gets you a third more hours and a worse rate.

That is not a trick. It is what happens when a higher tier is priced for features and seats rather than for volume, which is extremely common and worth checking on any tool before you upgrade. If you are upgrading for capacity, do the division first. If you are upgrading for a feature, fine — just know which one you are doing.

Note also that Descript counts media hours, not transcription hours: the allowance covers media you import or record, and transcription comes along with it. If you edit in Descript, the transcript is free in a sense, because you were paying for the editing anyway.

When does a subscription beat paying per minute?

HappyScribe publishes both, so the break-even is calculable rather than theoretical. At €0.20 a minute for credits:

Those crossing points come from simple division — €17 ÷ €0.20 = 85 minutes, and so on for each step up. The useful consequence: if you transcribe less than an hour and a half a month, do not subscribe to anything. Buy credits and stop thinking about it.

Most people who feel guilty about an unused subscription are in exactly that bracket.

Does the cheapest per hour mean the best choice?

No, and this is where per-hour tables mislead. These three are not the same product.

**Notta** is built around meetings and calls — it joins, records, transcribes and pulls out what was decided. The per-hour price is the lowest here by a wide margin, and the free tier caps a single recording at three minutes, which makes it a taster rather than a working tier.

**HappyScribe** is built around subtitles and translation, with human transcription available on the same platform. If your output is captioned video in several languages, that is a different job and the price reflects it.

**Descript** is a video and audio editor where the transcript is the editing interface — you delete a sentence in the text and it disappears from the video. Buying it for transcription alone is paying for a workshop to use one tool in it.

So the honest sequence is: decide the job, shortlist the tools that do that job, then compare per-hour prices within the shortlist. A comparison across the three columns above is a comparison of three different things.

There is a fuller side-by-side in which transcription tool.

What about human transcription?

HappyScribe lists human transcription from €1.75 a minute — around €105 an hour, against pennies for the machine version.

That gap looks absurd until you ask what an error costs. For notes you will read once and act on, machine transcription with a skim is obviously right. For a legal record, a medical note, or a quote you are about to publish under someone's name, a human is cheap insurance. The deciding question is never accuracy in the abstract — it is what happens when this is wrong.

A middle path that people forget: machine transcription plus fifteen minutes of your own correction gets you most of the way for the price of the machine. That works when you know the subject well enough to spot what went wrong, which for your own recordings you usually do.

What actually drives your accuracy

Before you pay more for a better engine, the cheapest improvements are all on your side of the microphone:

• A decent microphone beats a better model. This is the single largest factor and it is a one-off cost.
• One speaker at a time. Overlapping speech is where every engine falls apart.
• Say names and technical terms clearly the first time; most tools carry an error forwards once they have committed to it.
• Record locally rather than through a call platform where you can. OBS Studio is free and gives you a clean track.

People routinely upgrade plans to fix an accuracy problem that was caused by a laptop microphone in a room with a fan.

What I would do at three volumes

A couple of calls a month. Notta's free tier, or credits. Do not subscribe.

Weekly video with captions. Whichever editor you already use, with its captions — then the transcription is included in a thing you were paying for anyway.

Several hours a week, in more than one language. This is where a real subscription earns its place, and where the per-hour column finally becomes the right thing to look at.

The broader picture of what a content stack costs is in what AI tools actually cost per month.