Theo SidekickTurn your voice into polished text, in any app.
Free to startAll platforms
Used by teams at
Clinics
New Health Medical
Insurance
Lands in the apps you already use
Linear
Slack
Notion
Gmail
Telegram
Cursor
VS Code
Warp
Claude
ChatGPTTalking is the fastest way most people can get words out. The reason nobody writes that way is that raw speech reads like raw speech. Sidekick closes that gap, so the speed is usable.
Typing
40
words a minute
Talking
110 to 150
words a minute. Sidekick measures yours and shows it back to you.
What you say has fillers, restarts and the sentence you abandoned halfway. What lands does not, and it never loses a word you meant. Every dictation keeps both versions, so you can always see exactly what changed.
You said
hey so um can you actually wait can you tell the team that the the launch is gonna slip I think to like not Friday the following Monday because we are still waiting on uh legal to sign off on the the new terms page and uh yeah just let them know we will have a real timeline by by end of week sorry end of day Thursday
Theo typed
Can you let the team know the launch is slipping to Monday? We are still waiting on legal to sign off on the new terms page. We will have a firm timeline by end of day Thursday.
Follow one dictation from the hairline to the receipt. The bar on the left is the real Voice Bar, in the real order.
A hairline stands on the bottom edge of your screen. No window, no document. Your cursor is wherever you left it.
Capture starts before the key is fully down: a half-second ring already holds the syllable you began on. Beside it, Sidekick reads which app and which field you are in. Never a password field. Never the contents of a code editor.
A quarter-second post-roll catches your last word, so nothing you said is missing. Then the tone comes from where you are: casual in a chat, formal in an email. Fillers and false starts go. Two guards check the result, and one that dropped your words or echoed your screen is replaced by your own words.
Pasted at your caret, then verified by reading the field back. If the app will not take a paste, Sidekick types it instead. If nothing can take it, the words go to your clipboard and it tells you so.
The raw words, the finished text, the audio and how long each stage took, stored on this machine and nowhere else. Open History to play it back, see exactly what changed, or put the raw words back.
Prefer not to hold anything? Tap fn Space to lock it hands free. A single dictation can run for twenty minutes.
Plenty of tools can turn speech into accurate text. What happens next is where the day actually gets shorter: the tone, your vocabulary, the phrases you repeat, and what happens when something goes wrong.
A message to a friend and a message to a client should not read the same. Sidekick knows which app and which field it is writing into, and picks the tone to match. Five destinations, each one yours to override.
Styles
Messages
Work chat
Notes
Documents
Fix a name once and Sidekick remembers it. Client names, product names, the acronyms only your team uses: they get spelled right from then on, and every fix it learns comes with an undo.
Dictionary
Healthbird
heard as Health Bird
TheoVex
heard as Theo Vecks
Quito
heard as Keto
Shorthand turns a spoken cue into the text you type constantly. Your address, your booking link, the sign-off you use on every email. Say the cue, and the full version lands.
Shorthand
my email
ari@opencharts.com
sign off
Best, Ari
booking link
The full URL, every time
Speak any of them and Sidekick writes it down. Switch mid sentence and it follows. Pin the one you are speaking when you want the best possible accuracy.
Languages
Every dictation tool types what you say. These begin where that ends, and they are each one switch away.
It reads what you are answering.
Hold the reply key, say the gist of your answer, let go. Theo reads the message on your screen and writes the reply as you, in your voice, and it lands like any dictation with one undo. No other dictation tool knows what you are replying to.
The last 30 seconds, even though you never pressed the key.
Turn it on and Sidekick keeps a rolling half minute of audio in memory. Somebody says something worth keeping, you tap Rewind, and it is typed where your cursor is. The audio never leaves your Mac until that tap, and is dropped the moment dictation is switched off.
Dictate in a library, an open office, a sleeping house.
Ordinary microphone settings treat a whisper as noise and remove it. Whisper Mode turns that off and lifts the signal so a breath of a sentence is heard as a sentence. The bar shows an ear so you always know it is on.
Four presets, each a different writer. Turn on Auto and Theo picks one from the app your cursor is in, so a sentence spoken into a code editor and the same sentence spoken into an email do not come out the same.
Auto picks it in Cursor, Warp, VS Code, and every coding host
You said
make the login page remember me and not log people out every time they refresh
Theo wrote as Code Engineer
GOAL: Persist the session across refresh on the login flow. CONTEXT: Users are signed out on reload. REQUIREMENTS: 1. Store the session in a durable cookie. 2. Rehydrate on load before the first render. ACCEPTANCE: A refresh keeps the user signed in.
A plain-language idea becomes a precise brief: goal, context, numbered requirements, acceptance. Never an invented file name or library.
Auto picks it in ChatGPT, Claude, Gemini and the other assistants
You said
write me something about how I changed my mornings after burning out, for founders
Theo wrote as Prompt Engineer
Role: an editor writing for founders in their thirties. Task: an 800-word piece on how I restructured my mornings after burning out. Tone: conversational, not preachy. Format: prose, no listicle.
The rough instruction is restructured into role, task, context and format before it ever reaches the model.
Auto picks it in Mail, Gmail, Outlook, or anything that reads like an email
You said
tell Priya thanks for the deck and that the timeline slipped to Monday and we will confirm Thursday
Theo wrote as Email Master
Hi Priya, Thank you for the deck. The timeline has slipped to Monday; we will confirm the details by Thursday. Best, Ari
A ready-to-send email in your own voice, learned from the emails you have already written. Never a recipient you did not name.
Auto picks it in Docs, Notion, Pages, Obsidian, sixty words or more into a body
You said
so the thing about remote work is that it fixed the commute but it broke the hallway, all those small conversations
Theo wrote as Writer
Remote work solved the commute and broke the hallway. The small conversations that used to happen between meetings have nowhere to happen now, and much of what a team knows about itself used to live in them.
Spoken thoughts become essay prose. Your stance is kept, nothing is added.
Every preset is editable and teachable with your own writing samples. Add up to seven of your own, and bind any of them to a key: a tap rewrites what you selected, a hold dictates straight through it.
You get a tidy paragraph, and turning it into a ticket, a page or a message is still your job. Sidekick can do the other half, because it sits on a workspace that already connects to where your work lives.
Destinations run only after the words have landed at your caret. A send can never delay, replace or lose a paste.
Twelve words and a real cue, like a bug shaped like a bug or a note shaped like a note. Most dictations get no offer at all, on purpose.
Theo reshapes the words for that place and delivers them. The bar tells you exactly what happened, and the send stays on the History row if you would rather do it later.
Eight places the words can go
ScratchpadSidekick’s own
Always on hand, needs no account. The one destination on by default.
Theo NotesSidekick’s own
A note in your OpenCharts workspace, one tap to turn on.
Theo BoardsSidekick’s own
A task card on the board you choose.
Linear
Say the gist of a bug, and it becomes an issue.
Slack
Sent to the channel you already picked.
Notion
A long note becomes a page in your database.
Gmail
Drafted, or sent to your own inbox. Never to a third party.
Telegram
Say “remind me” and Theo messages you.
Only the Scratchpad is on by default. Connecting Linear for something else never quietly adds a button here. Nothing is sent to a person you did not name.
Sidekick has a second workspace. Hold the orb, hold a key, say “Hey Sidekick” or just type, and ask. Theo looks at the window you are in, answers out loud, flies the buddy to the exact control, and can take the steps for you, asking first.
Sidekick is in beta. Pointing, clicking and typing can miss, and it improves in every release. Transcription is fully supported. The Sidekick workspace is on the Mac only for now; Windows has Transcription in full.
Theo
Sidekick transcript
You
Theo
Pointed it out
Each one is a chip on the Theo card in the app, with a Try it that runs a real question so you can watch the card, the buddy and the voice do their part.
Each question takes one screenshot of the active window, so Theo answers about what you are actually looking at. Password, banking and sign-in windows are skipped on purpose.
Ask where something is and Theo finds the real button, field or menu on screen, then the buddy frames it with a neon ring while the answer plays.
Say what you want done and Theo gives the task its own card. Desktop control can bring apps forward and shares your mouse and keyboard; it is not an invisible second desktop. Ask him to email someone and the verified email workflow uses your connected mailbox or the local route you opted into in Settings > Access > Local email. Ask him to text someone and he uses Messages; in WhatsApp, Slack or Telegram he checks the person before typing. Sending waits for your go-ahead; email Autopilot applies only when your own words explicitly said send.
Questions about today, prices, products, people or news run a web search first and the sources appear on the answer card and under the answer in its conversation, so the answer is grounded instead of remembered. Say open the link, the second one or the Dell one and Theo opens that result in your browser; a Drive file or a document he found opens the same way.
Say remember and the fact is saved at once to your memory across Sidekick and Theo Chat, and written to the Theo remembers note in your Scratchpad. You can review and delete memories in OpenCharts.
Live conversation keeps listening and answering in real time, interrupts when you speak, and can glance at your screen every few seconds. It ends when you say goodbye or tap End.
Every key is yours to change, and the app keeps the ask-Theo keys apart from the dictation keys so the two never fire together.
Ask for a task and Theo says “On it.”, plans the steps, checks the screen after each one and shows every step on the card. Anything that deletes, sends or quits waits for your go-ahead, and when it has to stop it says why: the connection, a rate limit, the account’s credits.
Press the orb or your key while Theo is speaking and your new question takes over. Say “stop”, or tap the Stop pill, to end the answer, the pointing and anything Theo was doing for you. ⌘ ⇧ Esc sends the buddy home and closes every overlay at once.
A Theo Persona is a live video agent you build in OpenCharts: a face, a voice, its knowledge, its boundaries and the tools it may use, even your own likeness. From Sidekick, Talk opens it in a small window beside your work, and while you are on the call it can look at your screen, point, click, or carry a task through, then tell you what happened.
Coupon rules
Spring campaign · draft
The deck is on your screen. The minimum order is set right here, at forty dollars.
A small window at the edge of your screen: who is on the line, what they did, what they said, what they built. Sidekick comes onto your screen as the call connects, ready to point.
Ask what an error says or what a page is about; it looks through Sidekick and tells you.
Where is the export button? It shows you, or presses it for you when you ask.
Open an app, fill a form, reply to a message. Anything that deletes, sends or pays waits for your yes.
Say make a deck out of this page, or turn this into notes. The persona reads what is in front of you first, then builds from it, and the result lands as a card in the call window, one tap from OpenCharts. It also gets a quiet note as you move between apps, so a question about this page means something.
What it takes, honestly
A call from Sidekick runs as your own private call, even on a persona you published publicly. Steps that delete, send, pay or quit wait until you say yes. Hold your dictate key during a call and the microphone is Sidekick’s until you let go: the persona pauses and never hears what you dictate. The flask on the Personas card places a test call that bills no minutes.
Wispr Flow is very good at dictation, and we are not going to pretend otherwise. There are fourteen things a good dictation app does that they do well, twelve that Sidekick does and they do not, and six they still do better than us. Every row was checked against their own public documentation, and you can open all of them below.
14
things a good dictation app does, and they do them well
12
things Sidekick does that they do not
6
things they still do better than us
Spoken shortcuts that expand to saved text. Shorthand here, Snippets there.
Sends a dictation on to Linear, Notion, Slack, Gmail, Telegram, Boards or Notes. Eight destinations. Offered only when it fits, never sent without a tap.
Reads the message you are answering and writes the reply. Their screen context aids spelling; it does not write a reply.
Can transcribe entirely on your own machine. Optional, on Apple silicon. Theirs is cloud only, with no offline mode.
Your dictation is never used to train models, with no setting to find. Theirs is on by default for new accounts and must be switched off.
Transcripts and audio never uploaded to the vendor. Theirs stores both on their servers by default for cross-device features.
Cleanup and rewrites work together, chosen per dictation. Theirs turns cleanup off entirely while Transforms are on.
Four writing personas built in, and picked for you by app. Prompt Engineer, Code Engineer, Writer, Email Master. Theirs ships two, chosen by hand.
A free plan with room to actually use it. Thirty minutes a day here, roughly twenty-seven thousand words a week. Two thousand words a week there.
Included in a whole workspace, not a separate subscription. Presentations, notes, whiteboards, code and Theo, with dictation in the box.
More than a hundred languages. Twenty-four here.
An iPhone app on the App Store today. Ours is real and already in testing.
Runs on macOS 12 and 13. Sidekick needs macOS 14.
Compared against Wispr Flow’s public site and help center on 20 September 2026. Sources: wisprflow.ai, Help: What is Flow?, Help: Shortcuts, Why Flow, Help: Auto Cleanup, Flow for Android, Pricing, Help: Connection issues, Help: How to Use Transforms, Help: Security and privacy overview, Help: Improve the model for everyone & Private Cloud Sync. Their features change; if a row is out of date, tell us and we will fix it. Wispr Flow is a trademark of Wispr, Inc.
Dictation means handing a microphone to software all day, and asking Theo means letting it look, so it matters exactly where the audio goes, when a screenshot is taken and what is kept.
Every transcript and every recording is written to your own computer. They are never synced and never uploaded. The one exception is a Retry you tap yourself on a recording that failed.
On Apple silicon you can switch to an on-device engine and nothing leaves the Mac at all. It is optional, it covers 13 of the 24 languages, and cloud stays the default everywhere else.
If your cursor is in a secure field, dictation will not start. It says so instead of quietly recording.
Sidekick looks at the field you are writing in to pick a tone, and it withholds that context entirely in terminals, code editors, and anywhere the text looks like code.
A question takes one screenshot of the active window, at the moment you ask, and nothing is captured while Sidekick is idle. A live conversation may glance every few seconds while it is on. Password, banking and sign-in windows are detected and skipped.
Hey Sidekick is recognized on your Mac by the system's own speech recognition, and that audio never leaves the machine. Screen Recording, Accessibility and Speech Recognition are each yours to revoke in System Settings.
Sidekick keeps a quiet record of how much you have said, how fast you say it, and where the words went. It is your own data, computed on your machine.
Your usage
Last twenty weeks
Where your words go
Gmail, Outlook
Slack, Teams
Docs, Linear, Notion
iMessage, WhatsApp
One account, every device. Your dictionary, your Shorthand and your tone settings follow you between them.
Dictation everywhere, the ask-Theo workspace, and the on-device engine on Apple silicon.
macOS 14 or later · Apple silicon and Intel
Transcription only : the ask-Theo workspace (beta) is coming to Windows later.
Windows 10 or later · 64-bit
A keyboard you can talk to in any app, sharing your Mac's dictionary and settings.
Coming to iPhone
The iPhone keyboard is real and already dictating for a small group of testers. It arrives on the App Store with the same dictionary, Shorthand and tone settings as your Mac.
iOS 17 or later
Free with your OpenCharts membership. Signed, notarized and auto-updating, with no separate subscription for dictation.
Put your cursor anywhere, hold a key, and talk. When you need a hand instead of a typist, hold the orb and ask. Free with your OpenCharts membership, on the Mac and PC you already use.
