Hebrew transcription and meeting summaries

IvreetMeet is an Israeli service that turns Hebrew recordings into text and into a usable summary. You upload an audio or video file, and you get a full transcript split by speaker with timestamps, plus a structured summary with key points, decisions and action items. It was built for spoken Hebrew — interruptions, Hebrew-English code switching, names no dictionary knows — and the free plan covers 7 meetings and 120 minutes per organization each month.

What you get from one upload

How it works

1. 1Record as you already do. Zoom local or cloud recording, Google Meet, Teams, a phone recorder, or the browser recorder in IvreetMeet. Tell the participants that the meeting is being recorded and transcribed.
2. 2Upload the file. Audio or video up to 500MB: MP3, WAV, M4A, AAC, OGG/OPUS, FLAC, AMR, 3GP, WMA, AIFF, CAF, and video: MP4, MOV, WEBM, MKV, AVI. You can attach instructions for the summary — who is in the room, what the meeting is for, what matters.
3. 3Read the transcript. Each passage is marked with its speaker. Name the speakers once, fix anything that needs fixing, and search inside the text.
4. 4Use the summary. Key points, decisions and action items, written from the transcript rather than from memory. Not the emphasis you wanted? Write an instruction and ask for it again.
5. 5Export or share. PDF, Word or PowerPoint, or a share link. A public preview link shows only the first 20% of the transcript.

Why Hebrew is the hard case

A meeting is not dictation. People talk over each other, switch to English mid-sentence, and use names of people, customers and products that no dictionary contains. General models trained mostly on English handle clean dictation and fall over exactly here. IvreetMeet runs its own Hebrew speech model, self-hosted, and speaker separation is part of the system rather than an add-on. On the public ivrit.ai eval-d1 benchmark the model is level with the best public system (our own measurement, 17 September 2026); a new version only replaces the previous one if it makes fewer errors on the same real meeting recordings.

Privacy, in plain terms

• Your organization only. Meetings live in your account; members of your organization see them, nobody else does.
• No training on your meetings. Registered users' meetings are not used to train models.
• Voice recognition across meetings is off by default. A voiceprint is biometric data, so it is only stored for speakers whose consent was recorded. Unconsented voiceprints are deleted after 30 days, unused ones after 365 days.
• Deletion. A meeting, a speaker or the whole account can be deleted at any time.

The site in Hebrew

The product interface and the full guides are in Hebrew at ivreetmeet.co.il. These English pages cover the same ground for people who search in English.