Audio to text USA: what it costs, and who does it better
Transcrib Si transcribes audio to text with timecodes and speaker labels for $0 — three files a day, up to 30 minutes each, no card — or $10 a month when you pay yearly ($120 a year, or $20 a month). Every file is limited to 64 MB and 120 minutes on every plan. Below that, thirteen services Americans actually meet are compared with the prices they publish.
Our terms
| What | Value | Where it comes from |
|---|---|---|
| Free plan | $0: 3 files a day, up to 30 minutes per file, no card | Pricing page |
| Pro, paid yearly | $10 a month ($120 a year), 1,500 minutes of audio a month, DOCX export, REST API and MCP | Pricing page |
| Pro, paid monthly | $20 a month, the same allowance | Pricing page |
| Size of one file | 64 MB | Service contract |
| Length of one file | 120 minutes (two hours) | Service contract |
| Languages | Russian and English — set by you or detected automatically | Service contract |
| Export formats | TXT, Markdown, SRT, VTT, CSV, and DOCX on Pro | Pricing page |
| Speakers | named separately, on request | Service contract |
| Payment | a card through Stripe; we never see or store the card number | Pricing page |
The full plan list is on the pricing page. The API is metered separately: you pay for the audio you send.
What the other services publish, side by side.
The table below is built from the pricing and language pages of thirteen services, read on 22 October 2026. Click a column header to sort, and use the filters to narrow the list. Every row carries the source it came from and the date it was checked.
A cell that the research could not read says “not measured”. It never counts as a “no”, and it stays out of the filters. We did not buy a subscription anywhere: we compared published prices and published limits.
The short version of each one.
$10 a month paid yearly, or $20 a month. Three transcriptions a day for free, files up to 30 minutes. 98+ transcription languages, 134+ translations. Exports DOCX, PDF, TXT, SRT, VTT and CSV. Its pricing page sits behind Cloudflare, so we read the official render of 21 October 2026.
A meeting assistant from Mountain View and San Francisco. Pro costs $16.99 a month ($8.33 a month paid yearly). The free plan gives 300 minutes a month and three file imports. Six languages only: English, Spanish, French, German, Japanese and Chinese.
Pro at $8.17 a month paid yearly, with 120 free minutes a month, 50 uploads and a three-minute limit per session. 58+ languages. Trustpilot has hidden the company rating over a breach of the platform rules; 170 reviews are still there.
The widest language list in the sample: 150+. Basic is $8.50 a month paid yearly ($17 a month). A ten-minute trial, up to 45 minutes per recording. Human proofreading from $2.00 a minute on top of the machine price.
Pay as you go at $10 an hour, or Core at $25 a month ($275 a year). Thirty minutes free, no card. 54+ languages. A San Francisco company that says plainly that all prices are in USD and that your card provider converts them.
A New York and San Francisco speech-to-text API. Pay as you go: Universal-2 is $0.15 an hour and Realtime is $0.45 an hour. 185 hours of pre-recorded audio are free, with no card. 99 languages on Universal-2, and diarization costs $0.02 an hour.
When they choose us.
- Russian and mixed Russian-English speech. The engine was built for Russian and for the way real work switches between Russian and English mid-sentence. Every rival in the table covers far more languages; we cover two, and we cover them properly.
- The API is not a seat. The REST API and the MCP server come with the $10 Pro plan and are metered by the audio you send. Otter, Fireflies and tl;dv price their plans per seat, and their API access is a separate conversation.
- The limits are written down. 64 MB and 120 minutes per file, 1,500 minutes a month on Pro, the counter resets on the first of the month, and going over it returns an error instead of disabling the account.
- Six export formats, including CSV and Markdown. CSV carries one row per segment with time and speaker, which is what a script needs; Markdown is the same text as headings and lists.
- No card, no trial clock. The free plan asks for no card, so it cannot charge you when a trial ends.
Where we fall behind.
- Two languages against 98 to 150. TurboScribe, HappyScribe, Notta, Sonix, Transcribe.com and the rest transcribe dozens of languages. If your recording is in Spanish, French or Japanese, they are the right tool and we are not.
- No meeting recorder and no bot. Otter, Fireflies, tl;dv and Fathom join a Zoom or Meet call, split speakers and write notes. We take a file you upload. We do not pretend to be a notetaker.
- No PDF export. Our formats are TXT, Markdown, SRT, VTT, CSV and DOCX. Five of the services in the table publish a PDF as well: TurboScribe, Otter, HappyScribe, Sonix and Transcribe.com.
- The free plan counts files, not minutes. Three files a day is a lot of minutes and very few clips: three clips use up the day, however short each one is. Several rivals count minutes instead.
- No desktop app for Windows. The app is in the browser. There is nothing to install.
- Only a card, through Stripe. There is no ACH, no invoice and no US-local payment method on our side.
- Where the audio goes. Russian and mixed audio is processed by our own engine on infrastructure outside the United States. Pure English may be routed to a third-party cloud speech provider in the United States when that fallback is enabled on the server. If data residency in the US is a hard requirement, say so and we will tell you exactly what happens.
Five things people ask.
What does audio to text cost in the United States?
With us, $0 for three files a day of up to 30 minutes each, and $10 a month paid yearly ($120 a year, or $20 a month) for Pro. Across the market, the lowest published monthly price in the table is $8.17 (Notta, paid yearly), and by usage, AssemblyAI starts at $0.15 an hour and Deepgram at $0.0043 a minute. The table above gives all thirteen services with the price each one publishes, the source URL and the date it was checked.
Which US service transcribes the most languages?
HappyScribe, with 150+; then Transcribe.com with 120+, Fireflies with 100+, AssemblyAI with 99 and TurboScribe with 98+. We support two: Russian and English. If the language of your recording is not one of those, this table will tell you which of those services to look at.
Do you have a Windows desktop app?
No. Transcrib Si works in the browser, and the same service is behind the REST API and the MCP server, so a desktop client is a thin wrapper rather than a second product. Several rivals ship no official desktop app either.
Can I pay with a US invoice or ACH?
No. We take a card through Stripe, and we never see or store the card number. Yearly billing is $120, monthly is $20, and both can be cancelled from the Billing section of the app. There is no US-local payment method on our side, and we would rather say so than show a checkout that fails.
Where are the numbers on this page from?
Our own figures come from the service contract, the pricing page and our own measurements of 21 October 2026. Every other figure was read from the service's own public pages on 22 October 2026, and the source URL and the date sit in the last column of the table. Where a page could not be read, the cell says “not measured”.
Put a file in and read the text out.
Drop an audio or a video file below. Transcription runs without an account; sign in to keep the text, the history and the exports.
Drop an audio or video file
or — mp3, wav, m4a, mp4, up to 64 MB and two hours
Transcription runs without an account: sign in to keep the text, the history and the exports.
The Telegram bot is not live yet — the handle above is a placeholder, and we would rather tell you than send you to a dead link. The web widget above works today.
Related pages.
Transcription services in the United Kingdom
The same table for the British market, where most services still price in dollars.
Türkiye'de ses dosyasını metne çevirme
The Turkish market, with the Turkish-language rivals — and an honest word about what we do not do there.
Two services, one table, every difference named.
Transcribe Russian audio to text
What our own engine does with Russian speech.
The general guide, with the upload widget.
Our plans in full, including what fair use means.
Mark Burmistrov, Founder.
Every number on this page comes from the service contract, the pricing page or the public pages of the other services, and each one is dated. The author of the page and the publisher of the site are the same person. Questions and corrections: support@transcrib.si.