App directory · 43 apps
Not in one app. In the app you happen to be in. Press a key, say the thing, and the words arrive at your cursor — in the reply box, the ticket description, the commit message, the blank document you have been avoiding since this morning.
Read this first
Waver has no integrations with any app on this page. No API connection, no OAuth grant, no plugin, no bot. On Mac and Windows it pastes the finished text at your caret in whichever app is in front. In a browser or on your phone, you dictate in Waver and copy or export from there. Every name below is its owner’s trademark, used to tell you where to point your cursor — nothing more.
How the words get there
Which one you get depends on the device, not on the app you are writing in.
Press ⌥ Space on a Mac, Alt + Space on Windows, from inside any app. A small overlay opens and listens; press again to stop. The text is pasted where you were typing.
Your clipboard is put back afterwards. If the paste doesn’t land, the text stays on the clipboard and Waver says so — losing what you just said is the one unacceptable outcome.
The Flow Bar is hold-to-talk inside Waver: hold Control, speak, release. The cleaned text lands in the note, and you copy it into the tab next door.
Prefer a different key? ⌥, ⌘, ⇧ Space and ⌥ Space are all available in settings.
Long recordings, meeting transcripts and speaker-labelled notes come out as text you can move anywhere — including into apps that will never accept a keystroke from another process.
Notes sync across your devices, so the thing you dictated walking home is on your desktop when you sit down.
The directory
Find yours, and read what you’d actually say into it. If your app isn’t here it almost certainly still works — the list is examples, not a compatibility matrix.
08 apps
The eight windows most people have open right now. Between them they account for most of the typing in a working day.
The reply you've been rewriting in your head since Tuesday. Say the gist, run Turn into a reply, paste it in.
The same email with four more people on it — longer, which is precisely why dictating it saves more.
First drafts, where the shape of the argument matters far more than the sentences.
The cells everyone leaves blank: the status column, the feedback column, the notes field nobody fills in.
The meeting write-up you promised during the meeting. Talk it through once, paste the page.
The thought that usually dies somewhere between the car park and the front door.
A question you can only half-articulate. Prompt engineer turns the ramble into a role, a task, the context and a format.
Long background you'd never type out: the whole situation, spoken in one go, pasted as a single message.
07 apps
Where the blank page is the enemy. Speaking produces a bad first draft in three minutes, and editing down is a cheaper mental operation than building up.
Long documents where you already know the argument and dread only the opening paragraph.
An issue told the way you'd tell a friend across a table, then tightened with Polish.
The post that has been an outline for six weeks because outlines are not the hard part.
Body copy, plus the meta description you skip every single time.
A deliberately blank window. Dictate the one section you keep scrolling past.
Dialogue. Read aloud it lands much closer to how people actually talk.
Fast notes you'll search in a year — in whole sentences instead of three-word fragments.
07 apps
The writing around the code. Add your repo names and internal services to your Personal Dictionary once and they come back spelled your way.
Commit messages, and the comment that explains why rather than what.
Here the prompt is the work. Say the entire ask out loud, run Prompt engineer, paste the result.
Docstrings, and the migration note nobody reads until the exact moment they need it.
Release notes, and the TODO you'd otherwise abbreviate into something future-you can't parse.
PR descriptions and issue write-ups, with your service names transcribed correctly the first time.
A commit body longer than one line, without leaving the shell to get it.
Comments on a frame — the actual feedback, rather than the word 'hmm' and a pin.
07 apps
Every tracker has a description field, and every description field is empty. It is empty because typing it is boring, not because the context doesn't exist.
Issue descriptions with the reproduction steps actually in them, because saying the steps is faster than typing them.
Tickets, acceptance criteria, and the comment thread where the ticket is really decided.
Task descriptions long enough that somebody else could pick the task up.
Card backs — the part under the title that everyone leaves empty.
Status updates that are sentences rather than the words 'in progress' again.
Tasks with enough context that future-you knows what past-you meant by them.
Update columns, and the weekly summary nobody volunteers to write.
07 apps
Explaining something out loud is the test of whether you understood it. Dictating your notes runs that test for free, and Waver can turn the result into flashcards on a spaced-repetition schedule.
Daily notes, dictated straight into whichever file is already open.
Card fronts and backs spoken from memory — the recall test happens while you dictate.
Definitions in your own words, which is the only version that ever sticks.
Text boxes beside your handwriting, for when your hand has given up before the lecture has.
Lecture margins: what the lecturer meant, in your words, in the sixty seconds afterwards.
Notes on a paper while it's still open, before you forget why it mattered.
The prose between the equations — abstract, discussion, the paragraph you keep rewriting.
07 apps
Speaking is faster than typing, and the gap tells most in the places where you write the most words per day and deliberate over them the least.
Threads where one paragraph would land better than four consecutive one-line messages.
The post-call recap, written while you still remember what was agreed.
Long replies in a channel that expects long replies.
The message you'd send as a voice note — sent as text, for the people who never play voice notes.
The same, plus the note to self in Saved Messages that you'd otherwise not bother with.
The apology, the plan, the address — with the emoji spoken out loud and typed as emoji.
Comments and posts that sound like a person, on the grounds that one said them.
Why it holds up outside Waver
Once the text is pasted, it is just text in someone else’s app. So the work happens before it leaves.
Not a find-and-replace afterwards. Your product names, teammates and internal services are given to the transcriber before it hears a word, so it leans toward Kubernetes instead of guessing phonetically. Add terms, pin the important ones, and it learns from the corrections you make.
None, Light, Medium or High. None gives you exactly what you said, mistakes included — right for quotes, interviews and language practice. High gives you the paragraph you wish you'd written. The raw transcript is always kept, so “what I actually said” is one tap away.
Say a short cue and the whole block appears — your booking link, the paragraph you send four times a day. A cue only expands when it stands on its own, so mentioning the same words mid-sentence never drops boilerplate into a message you already sent.
“Fire emoji” becomes 🔥 in the text that lands. Small, and the difference between a dictated message reading as dictated and reading as written.
Auto-detected. You do not pick a language before you record, and switching mid-sentence doesn't break it. Live translation is there when your thinking and your audience don't share a language.
Two independent toggles, because they answer different questions. Set both and there is zero retention. Worth knowing before you dictate an unreleased feature name into a ticket.
Same recording, different shape
Four transforms ship built in. Point one at text you already have and it gets rewritten on demand — you are not dictating the same thing twice for two destinations.
Custom transforms are a name and a prompt, and nothing else. If you always turn notes into the same kind of standup update, write it once and it becomes a button.
What this page is not
Listed plainly, because finding out later is worse than reading it now.
There is no API connection, OAuth grant or webhook between Waver and any product named on this page. Nothing is written into your Notion workspace or your Jira project on your behalf.
The meeting recorder captures your microphone and your system audio on your own machine. Nothing joins the call as a participant, and nobody sees a bot appear in the attendee list.
Waver runs beside your editor, not inside it. There is no extension to install into VS Code, JetBrains or Xcode, and the desktop hotkey works the same in all three because it doesn't know which one it is.
Every product name here belongs to its owner and appears descriptively — to identify where you would be typing. None of these companies has reviewed, approved or endorsed Waver.
One last thing
Everything above is really the same instruction repeated forty-three times. Install the desktop build and press ⌥ Space — Alt + Space on Windows and Linux — from wherever your cursor already is. The overlay opens, you talk, you press it again, and the cleaned text appears in the field you were looking at. You never alt-tab to Waver, and Waver never needs to know which app it just typed into.
Inside Waver on the web or your phone it is a hold instead: hold Control, speak, release. A hold suits a sentence — there is no way to leave it recording by accident. A tap suits a paragraph, because holding a key for ninety seconds is its own small misery. Same key, both behaviours.
Not Fn, and this is worth being straight about: macOS eats that key at the system level to remap the function row and never delivers an event to a browser, an Electron window or a WebView. Tools that bind it ship a native helper with Accessibility permission. Waver offers keys that actually fire rather than a setting that quietly does nothing.
Free tier: unlimited transcription, 10 AI actions a month, no card, no ads. Premium from $4.13/mo. See the privacy policy before you dictate anything your security team would care about.
Questions
No — and that is the design, not a gap. Waver has no API connection, no OAuth grant, no webhook and no plugin in any product named on this page. On Mac and Windows the desktop build puts the finished text on your clipboard and sends a paste keystroke to whatever app is in front, so the words appear at your cursor. On iPhone and iPad you dictate inside Waver and copy or export from there. Nothing on this page implies a partnership, and no company named here has endorsed Waver.
Yes. Anywhere a caret blinks and Ctrl+V or Cmd+V pastes text, the desktop build can put words there. The list is forty-odd examples, not a compatibility matrix. The one real limit is apps that block synthetic paste — some password managers and secure input fields deliberately do, and they should.
On the desktop build it is Option+Space on a Mac and Alt+Space on Windows, registered globally so it fires from inside any app. Press it, a small overlay opens and starts listening; press it again to stop. Inside Waver itself — on iPhone and iPad — the Flow Bar is hold-to-talk on Control by default, with Option, Command, Shift+Space and Option+Space available in settings.
Because macOS consumes Fn at the system level to remap the function row and never delivers a key event to any browser, Electron renderer or WebView. The apps that do bind it ship a native event tap that needs Accessibility permission and a signed helper process. Binding Fn would be a promise Waver cannot currently keep, so it offers keys that genuinely fire instead.
No. The desktop build reads what was on your clipboard, writes the dictated text, sends the paste, then puts the previous contents back. If the paste doesn't land, the text is deliberately left on the clipboard and you are told to paste it yourself, rather than the feature failing silently and losing what you said.
Your Personal Dictionary is sent along with the audio, not applied as a find-and-replace afterwards, so the transcriber is already leaning toward Kubernetes rather than guessing phonetically. You can add terms yourself, pin the ones that matter most, and Waver learns from the corrections you make.
Only if you ask it to. Auto Cleanup has four levels — None, Light, Medium and High — and the raw transcript is always kept alongside the cleaned one, so you can restore what you actually said. None is the right setting for quotes and interviews; High is the right setting for a rambling memo you want to read as a paragraph.
120+ languages, auto-detected, written in their native scripts. You do not pick a language before recording, and switching mid-sentence does not break it.
Transcription is unlimited on the free tier, with 10 AI actions a month, no card and no ads. Premium starts at $4.13/mo and lifts the AI limit.
Free forever to start. Unlimited transcription, no card, no ads.