Every word,
timestamped and playable.

A transcript is only useful if you can find things in it. A four-hour meeting is tens of thousands of words, so EchoPilot’s transcript is built around navigation rather than volume.

Timestamps on every line, and tap to hear it

Every line carries the moment it was said. Tap it and the recording plays from there, so you check the text instead of trusting it and jump instead of scrolling. As the audio plays the current line is highlighted: a reading aid that tracks closely, not subtitle-grade synchronisation.

Search that behaves like search

Type a word and you get a match count, with next and previous stepping. Each match sits in its line with its timestamp, so a search term leads straight to the audio of someone saying it. What did we say about the budget, who agreed to the deadline, did anyone actually commit to Friday.

Who said what

On supported Pro meetings, EchoPilot identifies timestamped speaker turns, and you can rename or merge the labels for that meeting. Merging matters: a long meeting often splits one person into two speakers, and joining them is the difference between a useful transcript and an annoying one.

When reliable attribution is not available, EchoPilot shows the complete transcript without names and says so. It does not guess, and it does not drop the transcript. Voice Print, which would recognise the same person across meetings, is not available.

Correcting a misheard line

Names, jargon, anything said over a cough: correct the line directly. The original audio and the original text stay intact underneath as evidence, and any Meeting Minutes built from the old wording are marked out of date so you can rebuild those sections when you choose. Nothing rewrites itself behind you.

Long recordings, mixed languages

A Pro recording runs up to four hours and the transcript covers all of it. Audio is captured in segments as the meeting runs, which is why an interruption does not cost the recording, and the segments are assembled into one continuous document.

Speech that switches language mid-recording is transcribed as it was spoken. Separately, the Meeting Minutes translate on request into 32 languages, of which English, Arabic, Urdu and Hindi are verified end to end; the rest are best effort and the app says so. The transcript itself is never translated: it stays in the language that was spoken.

Hindi and Urdu are one spoken language written two ways, and speech recognition has to pick one. You can tell EchoPilot which one you read, once, in Settings. The minutes are then written in that language, in its own vocabulary and idiom, rather than the other one spelled in your letters. The transcript keeps the words as they were spoken, shown in the script you read. This is not the translation feature and it costs nothing extra.

Where you can read it

Transcripts are on iPhone, Android and the web app, on every device signed into your account. Capture happens on a phone, an Apple Watch or the Mac and Windows apps; the web app reads, searches, asks and approves, and never records.

A layer, not the product

Most tools stop here. EchoPilot treats the transcript as the evidence under everything else: the Meeting Minutes cite it, Ask quotes it with timestamps, and every action card can show the sentence that produced it.

What this does not claim

Transcript follow-along highlight. A reading aid that tracks the audio closely, not subtitle-grade timing.

Speaker attribution. Available on supported Pro meetings. When reliable attribution is unavailable the full transcript is shown without names, and the app says so rather than guessing. Voice Print is not available.

Transcript correction with rebuild. The original audio and the original text both stay intact underneath as evidence.

Minutes translation. 32 languages, of which English, Arabic, Urdu and Hindi are verified end to end. The rest are offered as best effort and the app labels them that way.