Usage¶

akou is one window with three columns: your calls on the left, the transcript in the middle, and Ask and your notes on the right. Everything here also exists on the command line and the local API, so what you do in the window and what your agent does land in the same call. See Agents and the command line.

A saved call: the sidebar lists the calls by workspace, the header shows three speakers with their talk time, the transcript has a time and a name on every line, and the notes sit on the right

The first run¶

The first time akou opens, it asks what you will use it for: Calls, Dictation or Both. It then asks only the steps that choice needs: where calls go, the permissions, the dictation key and engine, the speech models and the assistant. The models step downloads the speech models once, about 3.0 GB, and Record stays disabled with its reason until they are on disk. Run the setup again in Settings brings the same steps back. See Getting started.

Calls, grouped by workspace, newest first, with the live call on top. Each row shows the title, the day and the length.
Search calls finds a call by its title or workspace. It never searches what was said: that belongs to the system you hand calls to (Hand-off to your knowledge system).
New workspace adds a workspace. A call recorded into a new workspace also creates it.
Dictation, Models and Settings open their pages in the window. You leave a page from the sidebar.
The row at the foot says whether akou can record: Ready · Runs on this Mac, or what is missing, such as Models missing · Set up.

Recording a call¶

The Record row at the top holds everything a new call needs:

the workspace menu, which also offers New workspace…;
the call title (optional; you can rename the call later);
the Live: menu, which picks the model that writes the live transcript from the models on this Mac, and offers the Models page when there are none;
the Mic and Call meters, one for each side of the call;
Record, with its shortcut beside it. The default is Option+Command+R, and it works from any app; Settings changes it.

A call being recorded: the elapsed time, the microphone and call meters, Stop, the live transcript, a note already added and a second one being typed

While a call records, the row shows REC, the elapsed time, Stop, and buttons to mute your microphone and to pause. While the meeting app is in front, a small floating bar shows the time, the levels, Mute, Ask and Stop. It carries no transcript text, so it can stay up during a screen share; Settings turns it off (Floating bar while recording).

The first Record asks macOS for the microphone and for the system audio of the call. See Permissions.

The call header and the transcript¶

Over the transcript, the header shows the call's title, day, length, workspace and number of lines, and a chip for each speaker with their talk time. Click the title to rename the call, live or saved; the call's folder keeps its first name. Copy transcript copies the transcript so far as Markdown, and Share turns on a read-only live link to the call on your own network (off until you turn it on; share.bind in Configuration picks the network).

Every line of the transcript carries its time of day and its speaker. Lines still being spoken are drafts and change as more audio arrives; after the call, the final pass rewrites the transcript with its accurate model and labels the speakers again. While you scroll back during a call, Back to live returns to the newest line; on a saved call the same button reads Back to the end.

Right-click a line, or press Shift+F10 on it, for its menu: Play from here, Copy line, Copy with time and speaker ([15:41:07 Maya] ...), Name this speaker… and Fix this line…. A word you fix once on its line is fixed on every line of the call that has the same heard form, and a name or term you fix is learned for the workspace too.

Speakers¶

A new voice shows up as c1?, c2? and so on until you name it. Click a speaker's chip, or pick Name this speaker… on one of their lines, to name them; the same menu merges two speakers who are really one person (Merge into) and splits them again. Your own side is labelled with the name from Settings (Your name). An agent can do the same with akou name c2 Maya.

Notes¶

The note input sits at the foot of the right column. Type a line and press Enter: the note is stamped with the time you wrote it. Four markers set its kind:

Start with	It becomes
`-`	a bullet
`[]`	an action item, with a checkbox
`?`	an open question
`#`	a section heading

A note added by your agent (akou note, akou_add_note) lands in the same list, with a tag saying where it came from.

Ask¶

The box at the top of the right column answers questions about the open call, live or saved. Its menu holds ready questions: Catch me up, Was my name mentioned?, Decisions so far, Action items, and What did Maya say? for each named speaker. The answer shows as a card, with the times and speakers of the lines it cites ([15:41 Maya]), and the excerpts it was given under it.

An answer to a question about the call, citing the time of the line it rests on, with the excerpts it was given

akou builds a small context from the call on your Mac and sends only that to the assistant you picked in Settings: your Claude Code or Codex, a local model, or an API with your own key. With no assistant set up, the box becomes Search this call and finds the exact words, with their times. See Providers.

The player¶

When the open call has a recording, a player bar sits under the transcript: play and pause, the position, the speed, and a balance between your mic and the call. Play from here on a line starts at that line.

akou keeps an item in the macOS menu bar, the akou mark, with a red dot while a call records. Its menu stops the recording, starts a dictation, brings the window back, turns Open at login on or off, and quits akou. The application menu, akou beside the Apple menu while the window is in front, has Settings… and Install Command-Line Tool…, which puts the akou command on your PATH (The command line).

Models and Settings¶

The Models page lists the speech models on this Mac with their sizes, downloads the ones that are missing, and lets you cancel a download. The live models for the Record row's Live: menu, such as streaming Nemotron, are downloaded from here too.

The Settings page is plain rows grouped by topic, with a search box at the top. Every change saves on its own. There is no account to sign in to.

The Settings page: your name, open at login, the recordings folder, the workspaces, the call audio, the languages, the live transcript, the floating bar and the record shortcut

Every setting, including the ones the page does not show, is listed with its default in Configuration.