Your interviews. Your lectures. Your meetings. In your language.

You have the recording. You do not have eight hours to type it out.

Upload or record live. The words come back the way they were said - speaker by speaker, timestamped, searchable, and read right to left when that is the language. 20 in all: Arabic, Hebrew, Farsi, Turkish, Spanish, Portuguese, Hindi and 13 more.

Transcribe my first recording

Start free: 30 transcription minutes once per account, no card needed - just confirm your email. That is about 15 minutes of audio with automatic speaker separation.

Eligible new customers: 7 days free, then $8.99/month · Card required · We email 3 days before the first charge · Cancel in one tap

Typing it yourself 3-10 hours of work per hour of audio, by published estimates.

A transcription service Priced per minute quoted recording by recording, before anyone has heard it.

Here $8.99 a month $8.99 flat. Around 400 minutes a month with speakers separated - close to 7 hours. No credits, no tiers, no upsell.

An illustrative interview at a shared table, with a notebook and laptop.

Weekly team sync - recording

42:18 · today

2 speakers detected

0:000:26 excerpt · 42:18 total

This is the output format. Pick a language to see yours.

Which recording is on your desk?

Pick the one you are actually holding. Each page shows the output for that job, and the same way in at the end of it.

A Zoom, Teams or Meet recording

zoom recording transcription

The file you saved after the call, turned into text with the speakers split apart.

See the saved-recording workflow

A WhatsApp voice note

whatsapp voice note to text

Share it into the app from your phone and read it once, instead of listening to it four times.

See the voice-note workflow

Depth, user and customer interviews

interview transcription · user interview transcript

Every quote checkable against the audio before it reaches your deck, your readout or your paper.

See the interview workflow

A lecture or a study session

lecture transcription · class recording to text

Find the passage you need and replay that minute, instead of scrubbing the whole class again.

Students: half price for the first six months. See the lecture workflow

A recording for a case file

court recording transcription · hearing transcript

Prepare and review the draft yourself. This is not a certified transcript, and the page says so before it says anything else.

See what it can and cannot do

Research interviews and interpreter prep

user interview transcription · source transcript for interpreters

Source-faithful text with names and terminology supplied up front, exportable to wherever the work actually lives.

See the practitioner workflow

Most transcription tools stop at your language.

That is the whole reason this product exists. It is also the one thing here you can check somewhere else before you trust us.

What their own pages said on 11 September 2026

Otter's own help centre: "Otter currently only supports English, Spanish, French, German, Japanese, and Chinese (Simplified)." Not Arabic, Hebrew, Farsi, Turkish or Hindi - absent, not imperfect.
Zoom's own community manager: "Transcriptions are not yet available for the Hebrew language." The thread starts with a user getting "Unable to transcribe" and ends with the summaries arriving in the wrong language entirely.

Both links go to their pages, not to ours. Vendor language lists change - check theirs, then check ours. That is the point of putting a date on it.

Transcripts

20 transcription languages , detected automatically.
Right-to-left is native. Arabic, Farsi and Hebrew read the right way round, punctuation where it belongs.
Speakers separated, and correctable. Rename a voice once and it updates through the whole transcript.
The audio stays underneath. Every turn is timestamped, so a quote can be replayed rather than trusted.
One price. No per-minute meter, no tiers, no seats.

Outsmarting right to left, together

Right to left is not one problem. It is a hundred small ones, and most of them are local. So the product improves with you: your corrections, your names, your language.

Ben Zeevy What it heard in minute 6, and again in minute 41

Ben-Zvi You fixed it once - her name is right in every recording after

A word whose letters come back in the wrong order. An idiom your lecturer uses that no dictionary lists. A name spelled four ways in one conversation. English-first products are not going to fix those for you - there is no business case in it for them.

What already works today: in Hebrew, correct a word once and it leads on your next recording - a name, a dialect word, or a term you type in up front. In the other languages your corrections are kept in your library, on your device, tied to nobody else’s - they are not steering the engine yet. Hebrew went first because that is where the evidence was; the mechanism is the same for every language and each one is switched on against a measurement, not a hunch.

What we want from you: the one it still gets wrong. The word, and what it should have been. Send that and your language is what gets worked on next.

The person who writes this code reads every one of these, and answers. It is one developer, not a queue - which is the reason your language can move to the front of it.

Start improving it with your words

The first correction takes about four seconds, and it is yours - kept in your library, not pooled with anyone else’s.

The recording is made. Three questions decide what it was worth.

None of them are questions about us. They are the ones you end up answering yourself, usually at the worst possible moment.

Where is the rest of my week going?

An hour of two-voice audio takes most people four to six hours to type up: play, rewind, catch the word, rewind again. In Arabic or Hebrew, with a keyboard fighting the direction and punctuation landing on the wrong side, it takes longer.

Here the hour comes back as text you correct rather than text you build. Same attention, spent on the sentences that matter instead of on all of them.

Six hours of yours, or $8.99 for the month. Do that sum at your own rate - we would rather you check it than take our word for it.

Will I find the sentence again in three weeks?

The quote you need is somewhere around minute 50 of a file called REC_0043. Finding it means listening to the whole thing a second time, which is why most recordings quietly go unused.

Here the transcript is searchable and every turn carries a timestamp. Tap it and the original audio plays from that second, so a consequential sentence is checked rather than trusted.

Correct a word, rename a speaker, export the text - the source recording stays underneath all of it.

Is a draft enough for what I have to hand in?

Sometimes it is not, and that is worth knowing before you pay rather than after. There is no summary, no analysis, no findings and no certified transcript here.

What you get is a first draft with the audio still underneath it. If your case, course or client needs a certified or approved document, this is what you build it from - not the thing you hand in.

Try it on a recording you already have

One file you already have, in the language it was spoken - ten minutes tells you whether the text is usable. Your library is private to your account, encrypted in transit and at rest, share links expire, and you can delete a recording whenever you like. What we store, and who processes it

Companion for your calls. Text in your language.

Follow the conversation live, or keep a record for later.

Available now - Windows pilot

Live text in your language

Read Hebrew or English captions beside your call, translated into the other language. Finalized bilingual text reopens in your library on your phone.

Windows only. Hebrew to English or English to Hebrew. Recognition and translation can be wrong. Other live language pairs are still in research.

See pilot scope and setup Available today - Windows Companion

Record the call. Revisit the words.

Capture system audio and your mic - no meeting bot. After you stop, find a quote and replay it from the same library on your phone.

For customer calls, interviews and lessons you have permission to record. Set up Companion from your account.

Set up call recording

Transcription is normally sold by the minute, so a long recording is a decision. Here it is one monthly figure and you stop counting.

Pro - recording transcription

No tiers

$8.99 / month

800-minute monthly allowance: 6 hours 40 minutes with automatic speaker separation; up to 13 hours 20 minutes on Standard with speaker detection off

20 languages with correct right-to-left display

Automatic speaker separation and Best accuracy use the allowance twice as fast; rename speakers anytime

Cloud backup on every device + expiring share links

Continue to checkout

Start free: 30 transcription minutes once per account, no card needed - just confirm your email. That is about 15 minutes of audio with automatic speaker separation.

Eligible new customers only: 7-day trial with a 120-minute allowance · About 60 minutes with automatic speaker separation · Card required · Nothing charged for 7 days · We email a reminder 3 days before the first charge · Returning customers are charged $8.99/month immediately · Cancel anytime · Up to about 90 minutes per recording

Eligible trial customers get an email before the first charge. Everyone can cancel in one tap from the billing portal.

Common questions

Who gets the free week?

Eligible new customers get 7 days free, then $8.99/month. We email you 3 days before the first charge. Returning customers are charged $8.99/month immediately. Cancel anytime.

Why pay, when free transcription exists?

If a free tool already gives you text you would put in a document, in your language, with the speakers separated - use it. We mean that. The case for paying starts where that stops: a language the free tool does not really support, right-to-left text that comes back mangled, speaker turns you have to reconstruct by hand, or an hour of audio you are not allowed to paste into somewhere unknown.

Which languages are supported?

Arabic, Turkish, English, Spanish, French, German, Italian, Portuguese, Russian, Ukrainian, Polish, Dutch, Farsi, Hindi, Japanese, Korean, Chinese, Hebrew, Yiddish, and Amharic - plus automatic detection. Arabic, Farsi, Hebrew, and Yiddish display right-to-left.

Do you translate?

No. The transcript comes back in the language that was spoken, which is the whole point. One bounded exception is in pilot: on Windows, live captions can be read in Hebrew or English while the other language is being spoken. Other language pairs are still research.

How long can a recording be?

Record directly in the app or upload an existing file, straight from your phone's share menu. A whole meeting fits in one recording; longer conversations can be uploaded as more than one file.

What if I run out of minutes?

The 120-minute trial allowance does not reset. After a paid subscription starts, the 800-minute allowance resets at the start of each month. Transcription pauses at either limit; your existing recordings and transcripts stay available.

Clear support and privacy, before you sign up.

What the service stores, who helps deliver it, and how to reach us without sending sensitive payment information.

Privacy, in plain language

We store your account, usage and the recordings you choose to save. Third-party AI providers transcribe the audio, Cloudflare R2 stores the media, and Polar is the merchant of record - we never receive your full card number.

No advertising trackers, on the site or in the app.

Read the privacy page

Support, privacy, and billing help

Access, transcription, privacy, refund and billing questions all go through the same secure form. Include the account email we should reply to.

Never send card, bank, or identity documents through it.

Read the support page