Summarize a voice memo with AI without changing what was said

Smart Summary turns an audio recording into a clean, structured document: filler and false starts removed, broken sentences repaired, paragraphs, headings and lists added — while every fact stays exactly as it was spoken. sAInote, an AI voice-to-text app, does not paraphrase your recording into something looser; it edits it. Recordings up to 20 minutes are processed immediately on every plan, including as a guest with no account. Pro accounts can send recordings up to 2 hours, processed in the background. Nothing is invented, and nothing that carried information is dropped.

Summarize a recording 15 minutes of audio free, one time — no account and no password needed.

What you get from a Smart Summary

Facts stay as spoken: Filler, false starts and repetition are removed and trailing sentences are repaired, but numbers, names, claims and the degree of certainty you expressed stay exactly as you said them. Nothing is invented.
Structure, not just shortening: The output arrives as paragraphs, with headings and lists where the material has natural structure. A 40-minute recording becomes something a colleague can scan in a minute instead of a wall of text.
Long audio kept whole: Long recordings are split into sections, worked through in pieces and combined, so the last twenty minutes of a two-hour file get the same attention as the first. Speaker separation is included.
Edit & Translate in one pass: Inside Smart Summary, pick a target language and receive the edited summary translated in a single step. It translates the cleaned version rather than the raw transcript, and remembers your choice.
Pro Assistant for your own format: Write and save your own instruction — action items with owners, a client recap, five bullets — then apply it to any recording afterwards instead of re-explaining the format each time. Pro only.
Quick Note when wording matters: The verbatim mode is always available: the transcript as spoken, nothing added and nothing removed. It works on every plan, guests included, for anything you intend to quote.

How to summarize a voice memo with AI

1

Record or upload the audio

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

Choose Smart Summary

Select Smart Summary rather than Quick Note before processing. Recordings up to 20 minutes come back immediately; 20 minutes to 2 hours are processed in the background on Pro.
3

Set a target language if you need one

Still in Smart Summary, choose a language under Edit & Translate to get the edited summary translated in the same pass. Skip this step to keep the original language.
4

Read it, then find it later

Signed-in accounts save every note to a searchable history that matches both the transcript and the processed summary, across devices. Guest notes are not saved to an account.

What happens when you summarize a voice memo with AI

A summary is not a shorter transcript, and it is not a loose paraphrase either. Smart Summary works from the full transcript of your recording and rewrites its surface: filler words go, false starts and repeated half-sentences go, sentences that trailed off are repaired into grammatical ones, and the result is broken into paragraphs with headings and lists wherever the content has real structure.

What does not change is the content. Numbers stay the numbers you said. Names stay the names you said. A decision you described as tentative stays tentative. Nothing is added to make the text read better, and nothing that carried information is cut to make it shorter. That fidelity constraint is the point of the mode: the output should be something you can hand to someone else without listening to the recording again to check it.

This is where it parts company with a generic AI summarizer that compresses freely and rewrites in its own words. Loose paraphrase is fine for a podcast you half-remember and a problem for a client call. Smart Summary is available on every plan, including as a guest without an account, and it costs the same allowance as any other mode — a 12-minute recording uses 12 minutes.

Two-hour recordings are built up in pieces, not squeezed through one pass

Long audio is where summarizers quietly fail. The usual failure is truncation: the tool takes as much of the transcript as it can handle at once, summarizes that, and the final stretch of the recording collapses into a sentence or vanishes. You normally only notice when the decision made in the last ten minutes is missing from the notes.

sAInote handles long recordings hierarchically instead. The transcript is split into sections, each section is worked through on its own, and those results are then built up into the finished summary. Detail from the last twenty minutes of a two-hour recording survives the same way detail from the first twenty does. Long recordings also get speaker separation, so a discussion reads as a discussion rather than one undivided block of text.

The limits are firm and worth knowing before you press record. Anything up to 20 minutes is processed immediately, on every plan. From 20 minutes to 2 hours, processing happens in the background and requires Pro. Two hours is the hard cap for a single recording, and longer files are rejected rather than trimmed. Uploads are capped at 64 MB, in WAV, MP3, M4A, MP4, WebM, OGG or Opus.

When you should not summarize — use Quick Note instead

If you intend to quote it, do not summarize it. Anything where the exact wording carries the weight — a quote going into an article, a phrase from a negotiation, a commitment someone made in their own words, anything you want to keep on record — needs the verbatim text, not an edited version of it.

Quick Note is the mode for that. It returns the transcript as spoken, with nothing added and nothing removed, hesitations and repetition included. It reads worse and it is more faithful, which is exactly the trade you want when the wording itself is the substance. Quick Note is available on every plan, guests included.

You do not have to choose once and for all. A common pattern is to run Smart Summary to get something readable to circulate, and keep Quick Note as the underlying record for the moment someone asks what was actually said. For signed-in accounts, search across your note history matches both the raw transcript and the processed output, so a half-remembered phrase finds the note either way. Guests get no history: a guest note is not saved to an account.

When a plain summary is not the format you need

A general-purpose summary is a sensible default, not always the right shape. Pro Assistant covers the cases where it is not. You write your own instruction once — extract only the action items with owners, produce a client-ready recap, list every objection raised, keep it to five bullets — save it, and apply it to any recording from then on. It stays in your account, so a format you rely on every week is set up once. Pro Assistant is a Pro feature.

If you need the summary in a different language, Edit & Translate does it as one step rather than two. It runs inside Smart Summary mode, and the order of operations is the interesting part: it translates the cleaned, structured, edited output, not the messy raw transcript. Translating raw speech means translating the filler and the false starts as well, and the result reads like it. Editing first and translating second produces a document in the target language that was written as a document.

The target language you pick is remembered for next time, and translated notes are marked with a badge in your history. Roughly 35 target languages are available for content, and speech recognition covers considerably more languages than the 11 the interface itself is offered in.

Common questions about summarizing audio

Can AI summarize a voice memo without changing what I said?

Yes. Smart Summary removes filler, false starts and repetition and repairs broken sentences, but it does not alter the content: numbers, names, claims and the degree of certainty you expressed stay as spoken. Nothing is added to improve the text and nothing informative is cut to shorten it. If you need the literal words, use Quick Note.

How long can a recording be?

Two hours is the hard cap for a single recording. Up to 20 minutes is processed immediately on every plan, guests included. Between 20 minutes and 2 hours, the recording is processed in the background and requires Pro. Files longer than two hours are rejected rather than trimmed, and uploads are capped at 64 MB.

Can I summarize audio for free without an account?

Yes, 15 minutes of audio, one time. As a guest in the browser you get Quick Note and Smart Summary with no account, but no saved history. A free account also gives 15 minutes one time — not monthly — and keeps a searchable history. Pro gives 360 minutes, six hours, every month.

What is the difference between Smart Summary and Quick Note?

Quick Note returns the transcript verbatim; Smart Summary returns an edited, structured version of it. Quick Note keeps the hesitations and repetition and is what you want for quoting. Smart Summary is what you want for reading and circulating. Both cost the same allowance — a 3-minute recording uses 3 minutes in either mode.

Can I get the summary in another language?

Yes, through Edit & Translate, which works in Smart Summary mode only. It edits and translates in a single pass, and it translates the cleaned output rather than the raw transcript, so the result reads like a written document instead of transcribed speech. Roughly 35 target languages are available, and your choice is remembered.

What audio file formats can I upload?

WAV, MP3, M4A, MP4, WebM, OGG and Opus, up to 64 MB per file. You can also record directly in the browser, in the iPhone or iPad app, or on macOS using the Quick Capture hotkey — Control+Option+Command+N by default — without saving a file yourself first.

Can sAInote summarize a meeting while it is happening?

No. sAInote works on a recording, not a live call: there is no bot that joins a meeting and no real-time transcription during one. Record the meeting on your phone, your Mac or in the browser, then process it afterwards. A two-hour meeting fits inside the per-recording cap on Pro.

Is there an Android app or a browser extension?

No. sAInote, an AI voice-to-text app, runs as a web app in any browser, as native iPhone and iPad apps, and as a native macOS app with a global hotkey and a menu-bar item. There is also a REST API for developers. There is no Android app and no browser extension.

What happens to my audio after it is summarized?

Audio is held in private cloud storage — never public, never indexed, never shareable by URL — and cleaned up on a schedule after processing. Traffic through the REST API stores neither audio nor text. You can delete your account inside the app, which anonymises the data associated with it.

A summary you can pass on without re-listening

Smart Summary gives you a readable, structured document from a recording, with the facts left exactly where they were. When the exact wording matters instead, Quick Note is one choice away. Both are free to try for 15 minutes.

Summarize a recording