hmmmbeta
Download

AI notepad for meetings. Local. For macOS

Spots the meeting, transcribes it, writes up the summary and answers your questions

Download See how it works →
No bot, works in: Zoom · Meet · Teams · Browser
The hmmm window on macOS: a week overview with meetings and to-dos, folders and note search in the sidebar

From call started to finished notes, with nothing to do on your end. It notices the meeting, records it, tells the voices apart and writes the summary

Knows a call just started

Zoom, Meet, Teams or just a browser tab — hmmm spots the conversation and offers to record. Nothing to set up first

A video call grid with the speaker highlighted: hmmm notices a meeting has started on its own

Knows who said what

A live transcript with speaker labels. Your voice and everyone else is sorted into their own turns. Recognized right on your device

The Transcript tab in hmmm: turns split by speaker with timestamps, your own speech marked

Summary on the spot. And you can ask

Short takeaways and decisions land the moment the call ends. Forgot a detail? Just ask the meeting — hmmm answers from the transcript

The Summary tab in hmmm: the meeting goal and per-person action items, collected after the call

Not just a transcript, but notes: a block editor and AI skills for text

Dictation — everywhere you type. Fast voice to text in any app. Smart voice notes in hmmm

Dictation into any app

Works from a shortcut anywhere: mail, a chat, a search field. The text lands where your cursor is

Dictation into hmmm

In the app it arrives as more than text: a list of tasks becomes checkboxes, an enumeration becomes bullets

⌥SpaceInsert CancelEsc

Here's what's free. Forever

All of this runs on your Mac and costs nothing — no subscription, no time limit

Transcription by a local model

Recordings of your calls never leave the device

Your own keys to the cloud models

Connect models with your own API keys — for transcription, dictation, summaries and chat

Notes without limits

As many meetings and documents as you like, and search across them — on your Mac

Dictation

Dictate in any app. Smart dictation inside hmmm needs a cloud model for chat, or hmmm AI

Search across notes and meetings

Local keyword search. Ask AI about every meeting once a cloud model is connected

Integration

Connect Claude to your documents over MCP, with no cap on requests

Use an hmmm AI plan if you want it simple

Full access to everything in the app — no API keys, no VPN, no load on your memory

Models with no keys and no VPN

The best transcription and the best summaries. Reach the models without a VPN, without API keys and without spending your machine's memory

Speaker recognition

hmmm AI splits the transcript by speaker on its own and tags your own voice. No keys, no tedious setup

Search and chat across every document

hmmm AI knows the context of every meeting and note. It finds what was said and tells you what you might have forgotten

AI skills and dictation

Edit text with AI, dictate in any app, and get smart voice notes inside hmmm

Free$0forever
Plus$11.99/moearly price
Pro$29.99/moearly price
On-device transcription and dictation
✓
✓
✓
Your own keys for transcription, dictation, summaries and chat
✓
✓
✓
Unlimited notes
✓
✓
✓
MCP integrations with Claude
✓
✓
✓
iCloud sync across devices
✓
✓
✓
Smart voice notes
✓
✓
✓
hmmm AI — no keys, no VPN
—
✓
✓
Speakers out of the box
—
✓
✓
Search and chat across all meetings
—
✓
✓
AI text editing and formatting
—
✓
✓
Monthly cloud pool
—
10,000 credits≈ 28 hrs · 9 hrs with speakers
27,000 credits≈ 75 hrs · 25 hrs with speakers

hmmm AI is coming soon

There is no way to pay for it outside Russia yet. Leave your email and we will write to you the day it opens.

You choose how the app works

Pros
  • Everything on the device: the recordings and the transcripts
  • Works with no connection
  • Free and unmetered
Cons
  • Transcription quality is below what the cloud models manage
  • Uses up memory
  • No speaker separation
  • No summaries and no chat. Neither works without a cloud model
Pros
  • For transcription and dictation, the OpenAI, Deepgram and Nexara APIs are supported
  • For everything else, the Gemini, ChatGPT, Claude and Kimi APIs are supported
  • You pay the provider directly, and only for what you used
  • The key is yours — you can watch the spend and revoke it any time
Cons
  • You have to get a key and keep an eye on the balance
  • Where the provider is unavailable, it will not work without a VPN
  • The conversation goes to the provider’s servers
  • Two keys are needed: one for transcription and dictation, another for summaries and chat
Pros
  • Access to every AI feature with no keys, no VPN and no foreign card
  • Better quality and speaker recognition out of the box
  • Does not use up memory
  • Works from any country without payment or access problems
  • 7 days and 1,500 trial credits with no card
  • We store nothing. Neither recordings nor text remain on the servers
Cons
  • Paid. Two plans to fit what you need
  • The audio leaves the Mac to be processed by a provider. Same as with your own key

Short and honest

How is this different from Granola or Otter?

Theirs transcribe and summarize in the cloud — the conversation goes to their servers. hmmm transcribes on your Mac by default, can run on your own API keys, and puts the transcript into a block editor. The cloud is only needed for summaries and chat, and you can leave it off entirely

Which macOS do I need?

macOS 15 (Sequoia) or newer, ideally on Apple Silicon — the local models run on your Mac’s neural engine

What languages does it recognize?

Dozens, including English and Russian, with auto-detection

Do my recordings go to the cloud?

No — not until you switch the cloud on yourself. By default everything stays local, on your Mac

How far do the credits go?

Credits are spent on the cloud only. Plus gives 10,000 a month — roughly 28 hours of transcription; Pro gives 27,000, about 75 hours. Speaker labelling costs three times as much: about 9 and 25 hours. Local transcription and dictation are free and unlimited

Which models power hmmm AI?

Transcription and speaker separation run on Deepgram nova-3; meeting summaries and text editing on DeepSeek; the notes chat on GPT-5.6. On the free plan everything runs on your Mac. We swap models as better ones appear

Is it okay to record the other side?

Tell the others you’re recording — it’s polite, and required in many countries. From there it’s simpler than with cloud services: the recording and the transcript stay on your Mac until you connect a cloud yourself — your own key, or hmmm AI

Start using hmmm

Free, and no account needed. Requires macOS 15 or newer

Download the app