Transcribe and translate audio to another language

sAInote, an AI voice-to-text app, transcribes a recording and translates it into another language in a single pass — and it edits the text before translating, so the second language gets clean prose instead of a word-for-word copy of your false starts. The feature is called Edit & Translate and it runs in Smart Summary mode: filler removed, broken sentences repaired, paragraphs and headings added, facts left alone, and that cleaned version is what gets translated. You choose the target language once and it is remembered. The result is written text — not subtitles, not dubbing, not live interpretation.

Translate a recording No account needed to try it: 15 minutes of audio, once, in the browser.

What makes the translated version usable

Edited before it is translated: sAInote removes filler, repairs half-finished sentences and adds paragraphs and headings first. The translation is made from that cleaned text, so the second language never inherits the mess of the first.
One pass, not two tools: You do not transcribe in one place and paste into a translator somewhere else. Set a target language before processing, and the edited text arrives already translated in a single step.
Your target language is remembered: Choose the language you publish in once and sAInote keeps it for the next recording. Change it whenever the audience changes; the setting follows your account across web, iPhone, iPad and Mac.
Translated notes are flagged: Signed-in users get searchable history across devices, and anything processed with Edit & Translate carries a badge, so a translated write-up is easy to tell apart from a same-language one.
Speech recognition is wider than the interface: The interface itself comes in 11 languages, but speech recognition covers considerably more, so you can dictate in a language the menus do not use and still write the result up elsewhere.
Long recordings keep their ending: Long audio is handled with speaker separation and chunked summarising, so points made in the final ten minutes survive into the written output as clearly as the opening ones.

How to transcribe and translate a recording

1

Record or upload the audio

Record in the browser, on iPhone or iPad, or from the macOS menu bar with the Quick Capture hotkey Control+Option+Command+N. Or upload an existing file of up to 64 MB.
2

Choose Smart Summary

Select Smart Summary as the processing mode. Edit & Translate is offered here only — not in Quick Note, which stays verbatim, and not in Pro Assistant.
3

Pick the target language

Choose the language you want the finished text in. The choice is saved, so the next recording uses it automatically until you change it.
4

Process and read the result

Recordings up to 20 minutes come back straight away. Longer ones, up to the two-hour cap, run in the background on Pro and appear in your history when finished.

Why translating a raw transcript disappoints

Most workflows treat this as two separate jobs. A transcription tool turns speech into text, then a translation tool converts that text into another language. Each step does exactly what it promises, and the result still reads badly.

The reason sits between them. A verbatim transcript is an accurate record of speech, and speech is full of restarts, repeated words, abandoned clauses and sentences that change direction halfway through. A translator is faithful, so it carries all of that across. What you get is a precise translation of a mess, and someone has to rewrite it before it can be sent to anyone.

The damage also compounds. Broken grammar in the source gives the translation less to work with, so ambiguous phrasing gets resolved the wrong way and a stray filler word is treated as meaningful. Worse, the person tidying up afterwards usually cannot check the original, because they do not speak the first language — which was the reason for translating in the first place.

How to transcribe and translate audio to another language with sAInote

sAInote reverses the order. Edit & Translate runs inside Smart Summary mode: the recording is transcribed, then edited — filler removed, sentences repaired, paragraphs, headings and lists added, facts left unchanged — and that cleaned, structured version is what gets translated, in one pass rather than two.

Because the editing happens first, the target language receives finished prose. Nothing arrives with an "um" translated literally, and the structure decided during editing is preserved, so headings stay headings and lists stay lists. The output is text you can read, copy and paste into a document or message.

It is available in Smart Summary only, and that is deliberate. Quick Note is a verbatim record with nothing added or removed, so editing it before translating would contradict what it is for; Pro Assistant applies your own saved instruction instead. Once you set a target language it is remembered for the next recording, and every note produced this way is badged as translated in your history, so months later you can still tell which write-ups started life in another language.

What people use it for

The most common case is dictating in the language you think in and publishing in the language your team reads. You talk for five minutes in your own words, and what lands in the shared document is edited, structured text in the working language, without a rewriting round in between.

Recorded interviews and lectures are the second case. An interview conducted in one language can be written up in another, ready to quote from and edit. A lecture attended in a second language can be turned into notes in your first, which is a different task from translating a transcript — you want the argument in readable form, not every hesitation faithfully preserved. An English recording written up in Turkish, or a Turkish one written up in English, both work the same way.

Cross-border meetings are the third. Record the conversation, then process it once for the original language and once for the other side, so each audience gets edited, structured text rather than a raw transcript. This works on a recording you make and process afterwards; nothing joins a call and nothing is translated while people are still speaking.

Limits, file types and plans

Length is the limit that matters most. Recordings up to 20 minutes are processed immediately on every plan. Anything between 20 minutes and 2 hours is processed in the background and is available on Pro only. Two hours is the hard cap per recording; longer audio is rejected rather than truncated.

Uploads are capped at 64 MB, in WAV, MP3, M4A, MP4, WebM, OGG or Opus. Usage is counted in minutes of audio, and a three-minute recording costs three minutes whichever mode you use — translating it as well does not cost extra minutes.

Guests get 15 minutes of audio, once, in the browser, with Quick Note and Smart Summary but no saved history. A free account also gets 15 minutes one time — that allowance is not monthly — and adds history that syncs across devices. Pro gets 360 minutes, six hours, every month, plus Pro Assistant, Special Names and background processing for long recordings. Signing in is passwordless: a six-digit emailed code, a magic link, or Sign in with Apple or Google. Audio sits in private cloud storage, is never public or shareable by URL, and is cleaned up on a schedule after processing.

Questions about translating a recording

Can sAInote turn an English recording into Turkish text?

Yes. Record or upload the English audio, choose Smart Summary, and set Turkish as the target language. sAInote edits the English transcript first — filler removed, sentences repaired, headings and paragraphs added — and then translates that cleaned version, so the Turkish text reads as written prose rather than transcribed speech. The same works in the other direction.

Does it translate the raw transcript or the edited version?

The edited version. That is the difference from a transcribe-then-translate workflow: the transcript is cleaned and structured first, and the translation is made from that. A raw transcript carries every false start and repeated word into the second language, which is why two-step results usually need rewriting before anyone can use them.

Does translating change what I actually said?

No. Editing repairs how something was said, not what was said: filler is removed, broken sentences are repaired and structure is added, while the facts stay as they were. The translation is then made from that edited text. If you want an untouched, word-for-word record instead, use Quick Note, available on every plan.

Which languages can I translate into?

Roughly 35 target languages are available for Edit & Translate, and speech recognition covers considerably more languages than the interface does. The interface itself comes in 11: Arabic, Bengali, English, French, Hindi, Portuguese, Russian, Simplified Chinese, Spanish, Turkish and Urdu, with Arabic and Urdu rendered right to left. The target language is chosen separately from the interface language.

Can I use Edit & Translate with Quick Note or Pro Assistant?

No. Edit & Translate is offered in Smart Summary mode only. Quick Note is a verbatim transcript with nothing added or removed, so cleaning it before translating would defeat its purpose, and Pro Assistant applies your own saved instruction instead. Switch the mode to Smart Summary and the target-language choice appears.

Can sAInote translate a meeting live while it is happening?

No. sAInote works on a recording, not a live call, and no bot joins your meetings. There is no simultaneous interpretation, no dubbing and no subtitle file — the output is written text you can read, copy and edit. Record the conversation, then process it afterwards.

How long can a recording be, and what files are accepted?

Up to 20 minutes is processed immediately on every plan. Between 20 minutes and 2 hours is processed in the background and is Pro only. Two hours is the hard cap per recording; anything longer is rejected. Uploads are limited to 64 MB in WAV, MP3, M4A, MP4, WebM, OGG or Opus.

Are translated notes saved and searchable?

Yes, if you are signed in. Translated notes are saved to your history across devices and carry a badge marking them as translated. Search matches both the transcript and the processed output, so you can find a note using a phrase from either one. Guests get no history at all.

Do I need an account, and what does the free allowance cover?

No account is needed to try it. Guests get 15 minutes of audio, once, in the browser, with Quick Note and Smart Summary but no history. A free account also gets 15 minutes one time — not monthly — plus saved history. Pro gets 360 minutes, six hours, every month.

Start with one recording

Take a recording you would otherwise have transcribed, then pasted into a translator, then rewritten. Run it once through Smart Summary with a target language set, and compare what comes back with the two-step version.

Translate a recording