Libra Meet
In development · not yet released

LIBRA MEET · MEETING INTELLIGENCE

The notes are done before the meeting is.

Transcribes Korean, English and Chinese meetings live and translates them into the languages you choose. While the meeting runs, AI keeps rewriting the notes and records decisions and next steps as structured items.

See the capabilitiesContact sales
The screens below are a product concept with example content.
LIVE Speaker 2 · Glossary applied02:13
  1. Doyun KimKO

    배포 파이프라인을 이번 주에 스테이징까지 올리는 걸로 정리하죠.

    ZH我们就定这周把部署流水线推到预发布环境。

  2. Speaker 2EN

    I can take the staging rollout, but I need the config review done first.

    KO스테이징 롤아웃은 제가 맡겠습니다. 다만 설정 검토가 먼저 끝나야 합니다.

    ZH预发布上线我可以负责，但需要先完成配置评审。

  3. Doyun KimPartial caption

    그럼 설정 검토는 수요일까지

Product concept · example conversation

LIVE · TRANSCRIPTION AND TRANSLATION

Captions build as people speak. Final sentences get translated.

Takes a microphone or computer audio (tab or screen share) as input. Partial captions appear first; once a sentence is final it is translated into the target languages. Terms in the meeting glossary are applied to both transcript and translation.

AI CO-WRITTEN NOTES

Not a transcript dump. A document that keeps being revised.

As new sentences arrive, AI revises the whole page: merging duplicates, correcting earlier lines with what was said later, and updating the state of decisions and next steps. Lines you edited yourself are never overwritten.

meeting-notes.mdAI updated the notes just now

Deployment pipeline

Target for this week is staging. The config review was confirmed as a prerequisite.Evidence 02:14

AgreedShip to staging this weekEvidence 02:14
Finish the config reviewDoyun Kim · Wednesday
Run the staging rolloutSpeaker 2

CAPABILITIES

What one meeting leaves behind.

This is the scope built so far. Availability and pricing will be published at release.

01Transcription and input

Live transcription
Partial captions become final sentences; a dropped connection replays the gap on reconnect
Audio input
Microphone · computer audio (tab or screen share) · both at once
Languages
Korean, English, Chinese (Simplified); primary language plus experimental secondary-language detection
Speaker labels
Optional. Splits a turn when the speaker changes; names are assigned by people, never guessed from voice

02Translation and terminology

Live translation
Final sentences are translated; editing a sentence retranslates it
Glossary
Workspace and per-meeting terms with the translation to use in each language
Term correction
Optional correction of misrecognised terms against the glossary

03Notes and search

AI co-writing
Repeated whole-page revision during the meeting, user edits preserved, last revision undoable
Structured blocks
Next steps (owner, due date, state) and decisions (agreed / tentative / rejected), plus tables
Cited evidence
Each note and summary item shows the transcript sentence it came from
Semantic search
Searches notes and transcript together, combined with keyword search
Export
Markdown · text · JSON

CONSENT AND DATA

Not recording is the default.

Meeting content is sensitive material. Libra Meet is designed to keep only what it needs.

  1. 01

    No audio stored

    The audio used for transcription is not saved. What remains is text and notes.

  2. 02

    Consent recorded

    Consent is confirmed when the meeting is created, and the time and person are recorded.

  3. 03

    Access checks

    Only workspace members can open a meeting; only the host or an admin can delete it.

  4. 04

    Deletion scope

    Deleting a meeting removes the transcript, translations, notes and search index with it.

Telling participants and obtaining consent remains the responsibility of whoever runs the meeting. Disclosure duties for recording and transcription differ by region and by workplace policy.