SMeet
Sign in Download

New · Voice clone translation, 11 languages, on your server

Every meeting your company ever had, answerable in one line.

Calls are recorded, transcribed per speaker and indexed the moment they end. Ask a question in plain language and get the answer with the exact second it was said. Chat and HD video included, with the AI running on hardware you own.

Talk to us Get it on the Mac App StoremacOS

Also on the web and on Android

smeet · meeting memory

Why did we push the migration to next sprint?

Answer

Because the plan needed rewriting before the infra review, and the review lands on Thursday. Decided by Anna Jones on 12 March, in the planning call 1

2 recordings · 1 message · searched 14 calls

1Anna Jones12 March

“The plan needs rewriting before the infra review, so migration goes to next sprint.”

36:40
2David Park12 March

“Agreed — I will book the room for Thursday morning.”

39:48

0

recordings sent to a third-party AI service

99

languages transcribed, detected automatically

11

languages your own voice can speak

500 MB

per file, kept in your own object storage

How the memory gets made

Nobody takes notes. Nothing gets uploaded anywhere.

01 · The call happens

Recording starts with the first media stream if the group owner turned it on. Nobody is asked to approve a note-taker, and the recorder is your own service, running in your compose file rather than somebody else’s cloud.

REC 42:18auto, from the first stream
DP
SK
AJ
MR

02 · Every speaker gets a name

Whisper runs on your own machine and labels each line with the person who said it, timestamped to the second. Download as text or subtitles.

31:47Mike RossThursday works if the review happens in the morning.
34:20Sarah KimThen the notes drop the migration section entirely?
38:12Anna JonesI will own the revised plan and share it Thursday.
38:40David ParkPerfect, I will book the room.

03 · Ask it like a person

Search runs with your access rights, so an answer can never quote a room you were not in. Every claim links back to the second it came from.

When is the migration due, and who owns it?

Anna Jones, Thursday. Agreed at the end of the 12 March planning call.

38:12 Infra & SRE

After the call

The summary writes itself, with the receipts attached.

Decisions, owners and due dates come out of the same transcript, so nothing is invented and everything is quotable. Speaking time is there too, for the meetings where one person did all the talking.

  • Decisions, not bullet points

    Each one carries who decided it and the second it was said, so it can be reopened rather than argued about.

  • Owners resolved by voice

    The assignee comes from the speaker track, and when the name was never said the task asks the room instead of guessing.

  • Nothing published silently

    Below the confidence threshold a summary stays a draft. You confirm it, or it never enters the archive.

Release 4.2 planning

42:18

Team agreed to ship 4.2 on Thursday after the infra review. Migration risk moves to next sprint, Anna owns the revised plan.

Action items

  • AJRevised migration planAnna Jones · due Thursday
  • DPBook the infra review roomDavid Park · no date
  • SKCut the 4.2 release notesSarah Kim · due Wednesday

Who spoke

David Park38%
Sarah Kim27%
Anna Jones21%
Mike Ross14%

And the messenger it all sits on is a good one.

All features
  • Can you take the infra review?

    👍 2

    Taking it. Pinned the checklist.

    Messaging

    Threads, folders, pinned messages, reactions with your own corporate emoji, and an edit history nobody can quietly rewrite.

  • DP
    SK
    AJ
    MR

    Calls and conferences

    Start from any chat. Screen sharing, blurred backgrounds, per-person network bars, and guest links for people outside the company.

  • 0:47

    DELass uns das Infra-Review auf Donnerstag verschieben.

    Played back in Ivan’s own voice

    Voice, in any language

    Record your voice once. From then on every voice message can be heard in 11 languages, still in your voice, synthesised on your own hardware.

Your servers

The AI runs where your data already lives.

Transcription, summaries, search and voice synthesis are services in your own compose file, next to the database. No recording is handed to a third-party AI service, and there is no per-seat AI bill, because there is no vendor on the other end.

Talk to us See all features

docker compose up

smeet-backendbusiness logic, REST
smeet-graphqlAPI gateway, subscriptions
smeet-postgresmessages, threads, groups
smeet-miniofiles, avatars, recordings
smeet-redispub/sub, presence
whisper-svcper-speaker transcription
embedding + ollamasearch, summaries
tts-svcvoice clone translation

Everything persistent lives on the host under /data. Nothing is written inside a container.

See it running on your own hardware.

Start on our hosted instance today and move to your own servers whenever you want. Same app, same data model, nothing to migrate twice.

Talk to us See all features

  • macOS

    Native app, menu bar, dock badge

  • iPhone and iPad

    CallKit calls, VoIP push, full multitasking

  • Android

    Full-screen call notifications

  • Web

    Chrome, Safari, Firefox, no plugins

  • Windows

    Planned — use the web app

Self-hosted SMeet

Ask for the compose file and docs

Talk to us