Why every free transcription site caps you at 30 minutes (the maths)

A free transcription site caps you because every minute you upload costs it money on a server, and a visitor looking at ads does not pay for more than about five of those minutes. The cap is where the two lines cross. This post does the arithmetic with published prices, then explains why a site that runs the model on your computer instead of its own can skip the cap entirely.

What a free minute costs the site

Speech recognition is sold wholesale by the minute. The cheapest published rates in September 2026 are $0.0025 a minute at AssemblyAI, $0.0043 at Deepgram and $0.006 at OpenAI’s Whisper API. A site running its own servers pays less per minute at scale but has to pay for the servers whether anyone uploads or not, so the API price is a fair proxy for marginal cost.

Take the middle figure, $0.0043 a minute. A 30-minute file costs the site about 13 cents to transcribe. A 300-minute monthly allowance, which is what Otter gives free users, costs up to $1.29 per user per month if fully used.

Then there is the upload itself. An hour of MP3 at 128 kbps is about 58 MB. The site has to receive it, store it at least until the job finishes, and usually keep the transcript. Cloud storage and bandwidth are cheap per gigabyte, but a free user who uploads ten hours a month is half a gigabyte of traffic and storage the site did not get paid for.

What a free visitor earns the site

If the site shows ads, the standard measure is revenue per thousand page views, called RPM. Across niches AdSense publishers see page RPMs of roughly $3 to $12; a transcription site with business-software advertisers sits toward the upper half of that range. Call it $8. That is 0.8 cents per page view.

Now put the two numbers next to each other. Transcribing a 30-minute file costs about 13 cents. At 0.8 cents a page view, the visitor has to look at roughly 16 pages to pay for it. Most visitors to a tool page look at one or two.

A 30-minute cap does not make the free user profitable. It limits the loss to about 13 cents, which is a marketing cost the site can justify if a few percent of free users upgrade. Move the cap to 30 minutes per file and 300 minutes per month, and the maximum loss is about $1.30 per user per month, which is still fine if the paid plan is $17.

The caps, as published in September 2026

Otter: 300 minutes a month on the free plan, 30 minutes per conversation, three file imports for the lifetime of the account.
Rev: 45 AI minutes a month, one user, English only.
Descript: 60 media minutes a month.
Sonix: 30 minutes, once, as a trial.

Notice that the tighter limit is often not minutes but files. Otter’s three lifetime imports mean the free plan is effectively for live meetings, not recordings. That is not an accident. Live capture keeps you inside the product and converts better.

Why the numbers push everyone the same way

Once transcription costs money per minute, three things follow, and you see them on every free site:

A cap. Per file, per month, or both. The arithmetic above sets it.
An account. A cap only works if the site can tell who you are. Signup walls exist to enforce limits, not to be friendly.
An upsell at the moment of need. The cap triggers when the file is already uploaded and you are invested. That is the best possible moment to show a price.

None of this is sinister. It is what the cost structure requires. But it means “free” on a server-side transcription site always has an asterisk, and the asterisk is the cap.

What changes when the model runs on your machine

The same speech model that the cloud services use, OpenAI’s open-source Whisper, has been converted to run inside a web browser on the visitor’s graphics card through WebGPU. When a site does that, the cost line moves.

The site’s marginal cost per minute becomes zero. There is no upload, so no bandwidth. There is no server job, so no compute. There is no stored file, so no storage. The only thing the site serves is the model itself, about 200 MB once, which the browser then caches.

Redo the sum. A 30-minute file costs the site nothing. A three-hour file costs the site nothing. The visitor still earns about 0.8 cents per page view, and now every one of those cents is margin. There is no reason to cap, no reason to require an account, and no moment of need to sell into. Advertising alone can pay for the site, which is exactly the model this one runs on.

The cost has not vanished. It has moved to you, in the form of a model download and 30 to 40 minutes of your computer’s time per hour of audio. For most people that is a better deal than a cap, because the computer was going to be on anyway.

The trade-offs, honestly

Local transcription has its own asterisks, and they are different ones:

Your hardware sets the speed. A desktop with a discrete graphics card transcribes an hour in well under one; a thin laptop takes closer to the length of the recording.
It needs a browser with WebGPU, which today means desktop Chrome or Edge. Phones and Firefox are being worked on.
The models that fit comfortably in a browser are the small and medium ones. They are good on clear English speech and weaker on names and heavy accents than the largest cloud models.

What you get back is no cap, no account, no upload, and a recording that never left your disk. For lectures, interviews, meeting notes and first-draft subtitles that is usually the right side of the trade.

Frequently asked questions

Why not just charge less? Because the per-minute cost is real and the tools with the deepest margins already do charge close to it: the cloud APIs at 15 to 36 cents an hour. Consumer services sit above that because they pay for interface, storage and support.

Could an ad-funded site afford no cap on a server? Only if visitors averaged 16 or more page views per 30-minute file. Tool visitors do not.

Is a local model as accurate as Otter or Rev? On clean English audio the gap is small. On noisy audio or unusual vocabulary the large cloud models do better. Human transcription beats both.

Does “runs in the browser” mean my file is uploaded and deleted after? No. Nothing is sent. The processing happens in the tab, and you can watch the browser’s network panel stay empty while it runs.

Sources, checked 14 September 2026

Otter.ai, Rev and Descript pricing pages for free-tier limits
AssemblyAI, Deepgram and OpenAI pricing pages for per-minute API rates
AdSense page RPM ranges from publisher network reports collected in our research notes
MP3 size: 128 kbps × 3,600 seconds ÷ 8 = 57.6 MB per hour