Command mode turns spoken words like at, dot, slash, dash-dash, and pipe into symbols. Say "send" to press Return or "stop" to press Escape.
Private by design
After downloading a model, transcription runs locally on your Mac. No audio uploads, no account, telemetry, or cost.
Native to macOS
Otis can take over the key (F5) that normally starts Apple Dictation. Prefer fn or another key? Set your own.
Specifications
Mac
Apple silicon, M1 or later
macOS
Sonoma 14 or later
Processing
MLX on Apple silicon
Models
Cohere Transcribe, Whisper base to large-v3-turbo
Modes
Command, Dictation, Verbatim
Network
Model downloads only
Permissions
Microphone and Accessibility
License
Free and open source (*not public yet)
About
Otis is a macOS application that does one thing really well: talking to your computer 'in computerese', specifically talking to a model inside a harness, whatever agent you choose to pilot. Your words become intent that turns into action, a verbal human-computer interface.
The voice output on most dictation applications is just verbatim text. For Otis, I focused on how I would like dictation to work for me, a rather opinionated approach. And it's not necessarily done yet. I'm working on a small, filter-like transformer model that does what the current command mode's regex system does, but more dynamically, taking context into consideration. Right now, what Otis is good at doing is: you say “dash,” and you get -; you say “pipe,” and you get |.
For me, this is important because I want to talk to my computer, and I want the computer to understand what I'm saying. The shell doesn't understand the word “pipe,” but it understands the symbol. Otis is all about transforming the language of symbols into the actual symbols themselves. These symbols express intent, and intent is what leads you to action, which is the whole point of human-computer interaction.
Docs
sloppa docs work in progress
Otis is a menu bar app. It changes your speech into text. All processing happens on your Mac.
Download and install
Otis needs a Mac with Apple silicon (M1 or later) and macOS 14 or later.
Download the DMG, or run brew install --cask heavyetl/tap/otis.
Open the DMG. Move Otis to the Applications folder.
Open Otis. The Otis icon shows in the menu bar.
First launch
At first launch, Otis shows a setup window. The window has these steps: Microphone, Accessibility, Dictation key, Model.
Permissions
Otis needs two permissions.
Microphone. Click Allow Microphone. Then click Allow in the macOS dialog. Otis uses the microphone to hear your voice.
Accessibility. Drag the Otis card into the Accessibility list in System Settings. Or click Open System Settings and add Otis to the list. Otis uses this permission to see the dictation key and to type text.
Otis does not send your audio or text to the internet.
Choose a model
An automatic speech recognition (ASR) model is the software that changes speech into text. Otis recommends a model for the memory of your Mac.
Model
Parameters
Download
Languages
Whisper Base (English)
74M
150 MB
English
Parakeet TDT 0.6B v3
600M
1.3 GB
25
Cohere Transcribe (8-bit)
2B
2.4 GB
English
Recommendation by memory: less than 16 GB, Whisper Base. 16 GB to 31 GB, Parakeet. 32 GB or more, Cohere Transcribe 8-bit.
Click Show all to see more models. Otis downloads the model one time. You can change the model later in Settings.
Your first dictation
Otis sends text to the currently focused caret, wherever it is. You can even follow the data as it moves to it in select apps, works in all native ones.
Press and hold the dictation key. The default key is F5 (the microphone key).
Speak.
Release the key. Otis types the text at the cursor.
Recording keys
There are two ways to record.
Hold. Press and hold the dictation key. Release the key to stop. Otis types the text.
Toggle. Tap the dictation key. Otis records until you tap the key again. Otis types the text after each pause in your speech.
You can use these dictation keys: Fn (Globe), F5 (Mic), or a key of your choice. Change the key in Settings › Shortcuts.
If you use the Fn key, make these changes in System Settings:
Keyboard › Press Globe key to: set to Do Nothing.
Keyboard › Dictation › Shortcut: set to Off.
If you use the F5 key, set Keyboard › Dictation › Shortcut to Off.
Dictation modes
The mode controls how Otis changes your words into text. To change the mode, click the Otis icon in the menu bar and select a mode. Or hold the dictation key and press the mode key. The default mode key is 5.
Command
Use this mode in a terminal. Otis changes symbol words into symbols. Otis changes number words into digits. Otis joins spelled letters into one word. Otis makes the first letter lowercase and removes the period at the end.
Example: say L S dash L A pipe grep dot swift. Otis types ls -la | grep .swift.
Dictation
Use this mode for normal text. Otis types your words as spoken. Otis obeys the Submit and Cancel commands only. Otis does not change symbol words.
Verbatim
Otis types the text exactly as the model heard it. No commands are active.
Voice commands
Commands work in Command mode and Dictation mode. Symbols and numbers work in Command mode only.
Submit
Say send, submit, or enter as the last word. Otis removes the word and presses Return.
Cancel
Say only stop, escape, or cancel. Otis presses Escape. Otis does not type text.
Symbols
Say the name of a symbol. Otis types the symbol.
dot, period
.
comma
,
colon
:
semicolon
;
question mark
?
bang
!
quote
"
single quote, apostrophe
'
backtick
`
open paren, close paren
()
open bracket, close bracket
[]
open brace, close brace
{}
slash
/
backslash
\
pipe
|
double pipe, logical or
||
ampersand
&
double ampersand, logical and
&&
equals
=
double equals
==
not equals
!=
plus
+
dash, minus, hyphen
-
dash dash, double dash
--
underscore
_
less than, redirect in
<
greater than, redirect out
>
append redirect
>>
hash, pound sign
#
dollar
$
percent
%
caret
^
asterisk, star
*
tilde
~
at
@
space
one space
Numbers
Say number words from zero to nine. Otis types digits. Example: say one two three. Otis types 123.
Spelled words
Say two or more single letters. Otis joins the letters into one word. Example: say L S. Otis types ls.
Settings
Click the Otis icon in the menu bar. Select Settings.
Microphone
Select an input device from the list. Or select the system default. The menu bar also has this list.
Model
Select a model from the list. The list shows the download size and the status of each model. Otis downloads a new model one time. The menu bar icon shows the download progress.
Shortcuts
Set the dictation key and the mode key. Select a preset or click Set your own and press a key.
Open at login
Turn on Open at login. Otis starts when you log in to your Mac.
Updates
Select Check for Updates to get the latest version.
Troubleshooting
Otis cannot hear or type
Click the Otis icon. If the menu shows Permissions needed, click Setup. Follow the steps to give the missing permission.
If you installed a new version, the old Accessibility entry can stop working. Open System Settings › Privacy & Security › Accessibility. Remove Otis from the list. Then add Otis again.
Otis cannot type when another app has secure input on. Password fields, sudo, and the iTerm Secure Keyboard Entry option turn secure input on. Otis makes a beep sound. Turn secure input off, then try again.
The recording key does not work
Make sure Otis has the Accessibility permission.
If you use the Fn key, set Keyboard › Press Globe key to to Do Nothing. If you use the Fn key or the F5 key, set Keyboard › Dictation › Shortcut to Off.
If another app uses the same key, set a different key in Settings › Shortcuts.
The wrong microphone is active
Click the Otis icon. Select the correct input device from the microphone list.
A model does not load
The menu shows Couldn't load and the name of the model. Otis continues with the last model. Check your internet connection. Then select the model again.