Private voice typing for Mac

Say it once.
It’s already typed.

Hold Right Option, talk the way you actually talk, and finished text lands where your cursor already is. The speech model and the cleanup model both run on your Mac.

  • Free & MIT-licensed
  • No account, ever
  • Apple Silicon · macOS 14+
Messages Message mode

Filler out. Self-correction applied. Your meaning intact.

Heard: “tell Maya um I can do three, actually make that three thirty”

Tell Maya I can do 3:30.

Ready 0.0s

The mechanic

One key. No new app to learn.

Velora never asks you to say “new line” or “comma.” You hold a key, you talk, you let go. Everything between those two moments happens on your machine.

  1. Hold

    Right Option by default, or any shortcut you prefer. A small pill appears near your cursor — it never steals focus from the field you are typing into.

  2. Talk

    Ramble. Pause. Change your mind halfway through a sentence. Velora is built around the way people actually speak, not the way dictation software wishes they did.

  3. Release

    Cleanup runs locally, the text is pasted at your cursor, and it also stays on the clipboard — so nothing is lost if the app you are in refuses an automatic paste.

Destination-aware

The app you’re in decides the format.

Velora reads which app is in front and switches its formatting rules. Slack gets a message. Mail gets paragraphs. Your editor gets your identifiers back verbatim, with no trailing period to break the command.

Message

Slack · Messages · WhatsApp

“are we still on for thursday i can move things around if not”

Are we still on for Thursday? I can move things around if not

Terse. No invented greeting. A short line keeps its missing full stop.

Email

Mail · Outlook · Spark

“thanks for sending that over i read it last night separately we should talk about the pricing page before friday”

Thanks for sending that over — I read it last night.

Separately, we should talk about the pricing page before Friday.

Paragraph breaks where the topic shifts. Never invents a sign-off you didn’t say.

Note

Notes · Obsidian · Bear

“first pull the logs second check the retry counter third file the ticket”

1. Pull the logs
2. Check the retry counter
3. File the ticket

Spoken ordinals become a numbered list. Plain prose stays prose.

Code

VS Code · Zed · Terminal

“git commit dash m fix the retry backoff”

git commit -m "fix the retry backoff"

Verbatim. Nothing capitalised, no trailing period — a period breaks a shell command.

Beyond dictation

The parts you only notice when they’re missing.

Private meeting memory

Records the room. Never the screen.

After you confirm — every time — Velora captures microphone and computer audio, keeps a visible recording indicator running, and writes a transcript and notes you can search locally. It captures audio only.

AirPods, unharmed

Your music pauses itself.

For direct dictation, Velora pauses the one active media source — including browser playback — and resumes only what it paused. Meeting capture leaves call audio running.

Voice edit

Select it. Say what to change.

Highlight a sentence, describe the edit out loud, and it is rewritten in place. If the result is wrong, ⌘Z puts it back exactly as it was.

Languages

99 languages, one model.

The default speech model is Whisper large-v3-turbo — multilingual, including Hindi and Indian English. Set a language or let Velora detect it.

Personal dictionary

It learns your nouns.

Teach it the names, products, and jargon it keeps mangling. Sync is optional, uses your own iCloud Drive, and carries confirmed terms — never audio or transcripts.

History

Nothing is thrown away.

Every dictation keeps its raw transcript alongside the cleaned result, searchable on your Mac. If cleanup got it wrong, the original is still there.

Private by architecture

Not a policy. A missing network call.

Most dictation tools promise privacy in a document you have to trust. Velora’s speech recognition and cleanup have nowhere to send your audio, because the models are sitting on your disk. The source is on GitHub if you would rather check.

Read the privacy model
Your voice
stays on the Mac
Transcripts & history
stay on the Mac
Meeting audio & notes
stay on the Mac
Accounts, telemetry, ads
do not exist
Model download, first run
uses the network once
Update checks
optional, can be turned off
Dictionary sync
optional, your iCloud Drive

Alpha Velora for iPhone

An early experiment, not a product yet.

The iPhone companion turns the Action Button into a voice key: press, speak, finish, and the transcript is on your clipboard. Recognition is on-device, using Apple’s own speech models.

It is alpha and it is not stable. Expect rough edges, bugs, and breaking changes between commits. It is not on the App Store, there is no TestFlight build, and it is not something to depend on. Build it only if you are comfortable running unfinished software from source.

Install

Free to install. Yours for good.

Download the signed, notarised app or install it with Homebrew. On first launch Velora asks for microphone, input-monitoring, and accessibility permissions, then downloads the models it needs. Setup takes a few minutes and needs the internet once — after that, dictation is local.

Terminal
$ brew install --cask sushilk1991/tap/velora

Velora is ready in Applications

Straight answers

Before you install.

Does Velora ever touch the network?

Models download once during setup, and optional update checks contact GitHub. During dictation, speech recognition and cleanup run locally — your audio and transcript are not uploaded. Update checks can be turned off.

What Mac do I need?

An Apple Silicon Mac running macOS 14 or later. Disk use depends on which speech and cleanup models you pick; Velora chooses a cleanup tier based on how much RAM you have.

Is it really free?

Yes. It is MIT-licensed and open source, with no account, no subscription, and no paid tier. There is nothing to upsell you because there is no server to pay for.

Is the iPhone app in the App Store?

No — and it is worth being blunt: the iPhone app is alpha and unstable. The Xcode project is in the repository and builds on iOS 17 or later, but it is unfinished, changes without warning, and is not something to rely on. The Mac app is the real product.

How do I get help?

Email sushilk.1991@gmail.com or open a GitHub issue. Velora is pre-1.0 and actively developed; bug reports genuinely help.

Free. Open source. Local.

Stop typing. Keep writing.

Already using Velora? A star is how the next Mac user finds it.