Speechy vs VoiceInk

VoiceInk is a great answer to "why pay for dictation?": it is open source, runs on your Mac and costs $25 once. Speechy also keeps your voice on your Mac. The difference is everything around the words. Here it is, with real examples.

At a glance

	Speechy	VoiceInk
Where your voice is transcribed	On your Mac	On your Mac, or a cloud service with your own key
Your words while you speak	Yes, with Parakeet	Yes, with streaming models
Two languages in one dictation	Yes, detected after each pause	Language is set per mode
Ignores other people's voices	Yes, with Only My Voice	No, it is an open request
Claude Code and Codex	Questions, permissions and plan limits in your notch	No agent inbox
Price	Lifetime from $99, or Pro at $12.99 a month	$25 once for 1 Mac, free if you build it yourself
Needs	macOS 14 or later, Apple silicon	macOS 15 or later, Apple silicon for local models

Only your voice gets typed

This is the one VoiceInk users keep asking for. A "speaker lock" request is open on its GitHub, and earlier requests to tell speakers apart were closed as not a focus.

Speechy has it today. Read one sentence aloud for about 10 seconds and turn on Only My Voice. Speechy then leaves out words from the colleague behind you, the TV or a speakerphone, and keeps learning your voice as you use it. Someone talking over you at the same moment can still slip through, but everyday background chatter does not.

Only My Voice: read one sentence aloud and Speechy learns your voice

Switch languages mid-dictation

In VoiceInk, the language belongs to a mode, so you set it per app or per shortcut. In Speechy, pick your languages and turn on Multiple languages in one dictation. Speechy works out the language again after each pause, so you can move between languages without changing anything.

You say“रिपोर्ट तैयार है। the report is ready”

Speechy typesरिपोर्ट तैयार है। The report is ready.

With Hindi and English picked. Speechy mixes English, Spanish, French, German, Italian, Portuguese, Dutch, Hindi, Japanese and Chinese.

Speechy's language settings, with Multiple languages in one dictation

Built for Claude Code and Codex

VoiceInk can send your text through a coding CLI to clean it up, but it has nothing for answering your agents. Speechy puts Claude Code and Codex in your Mac's notch:

Answer without switching windows. When an agent asks a question, wants permission to run a command or shows you a plan, it lands in the notch. Press 1 to allow or 2 to deny, J and K to move between them and E to clear one. Answered too fast? Esc takes it back within 2 seconds.
Watch your plan limits. Your Claude and Codex usage sits beside the camera, read straight from the tools, with no keys to paste. The Usage tab shows the 5 hour and weekly windows.
Talk to an agent by name. Hold the key and say "api, run the migration again on staging", and it goes to that project's agent.
Read replies right there. Replies show formatted, with tables and code. Drop in up to 6 images, or drag the corner to make the chat bigger.
Name files out loud. Turn on Reference coding files, and the file names you say become @file references in Claude Code for VS Code.

A Codex permission request in Speechy's notch, with Allow and Deny

Claude and Codex plan usage beside the camera

Talking to the api agent from Speechy's notch

Everything else Speechy does

Here is the rest of Speechy, with real examples. Every "Speechy types" box is the app's own output for that sentence, taken from Speechy's test suite. All of it runs on your Mac, with no AI provider or API key to set up.

It cleans up the way people talk

You say“so I think we should, you know, ship it on Monday”

Speechy typesSo I think we should ship it on Monday.

You say“we need to um we need to fix the login bug”

Speechy typesWe need to fix the login bug.

You say“I wanted to ask you about the, I wanted to ask you about the budget for next quarter”

Speechy typesI wanted to ask you about the budget for next quarter.

It keeps up when you change your mind

Say "sorry", "I mean", "make it" or "or rather", and Speechy swaps the old time, date, amount or name for the new one. "Make it work" and "sorry I'm late" are left alone.

You say“let's meet at 8pm at the library, sorry make it 9pm”

Speechy typesLet's meet at 9 PM at the library.

You say“ask Priya about the launch date, I meant Ravi”

Speechy typesAsk Ravi about the launch date.

Lists, without touching the keyboard

Count your points out loud, or say "bullet", and Speechy writes the list. In Mail, Notes, Pages and Word the dashes become bullets.

You say“number one call Sam number two send the deck”

Speechy types1. Call Sam 2. Send the deck

You say“we need three things: a budget, a timeline, and an owner”

Speechy typesWe need three things: - A budget - A timeline - An owner

You say“things to buy bullet milk bullet eggs bullet bread”

Speechy typesThings to buy - Milk - Eggs - Bread

Numbers, money and times, written properly

You say“the total is twenty five thousand dollars”

Speechy typesThe total is $25,000.

You say“the meeting is at seven pm”

Speechy typesThe meeting is at 7 PM.

Punctuation and new lines when you ask

You say“send it new line thanks”

Speechy typesSend it Thanks.

You say“hello world question mark”

Speechy typesHello world?

Emails come out shaped like emails

Start with a greeting and end with a sign-off, and Speechy puts each on its own line. It skips this in chat apps, terminals and code editors.

You say“hi Sam, let's meet at 8pm at the library, sorry make it 9pm, cheers Jay”

Speechy typesHi Sam, Let's meet at 9 PM at the library. Cheers, Jay

Your words, spelled your way

Add names and jargon to your dictionary, fix a word Speechy keeps mishearing with a replacement (in every app, or only in one), and save snippets for things you say often. Say "insert" and the snippet's name.

You say“insert my address”

Speechy types42 Main St.

With a snippet called "my address" saved.

Speechy's dictionary: words, replacements and voice snippets

It spells what is on your screen

With App Context on, the names in the window you are dictating into become spelling hints. It stays on your Mac, password fields are always skipped, and you can leave out any app.

You say“look at dictation controller dot swift”

Speechy typesLook at DictationController.swift.

With DictationController.swift on screen.

Context you control: the names Speechy used from your screen, and the apps it leaves out

Undo a dictation, or throw one away

Press ⌘⌥Z within five minutes and Speechy removes the last dictation and puts back anything it replaced. Press ⌘Esc while talking to drop the take, so nothing is typed or saved.

Speechy's shortcuts: Fn to dictate, ⌘⌥Z to undo, ⌘Esc to discard

Paste, type or just copy, and send

Speechy pastes by default, then puts your clipboard back the way it was. It can type the words out instead, or leave them on the clipboard. Turn on Auto submit and it presses Enter (or ⌘Enter) for you.

Speechy's output settings: paste method, clipboard and auto submit

Pick your look

Speechy shows a small pill while you speak: a compact pill, a wide pill with your words, a Dynamic Island style bar or a one-line ticker.

Speechy's four recording styles: Compact, Wide, Dynamic Island and Ticker

And more

Music pauses while you talk, then carries on, with Background audio set to Pause media.
Help when the mic is wrong: if Speechy hears nothing, it plays your recording back and opens the microphone picker in one click.
Stats for your speed, streaks, golden hour and time saved, and a Speechy pass you can share.
Light on memory: models free about 1.2 GB after 15 minutes idle and reload as you start talking.
A whole email from one sentence with Smart Polish on: "draft an email to Sarah saying the deck is delayed to Thursday".
Bring your dictionary from Wispr Flow, Superwhisper, Monologue, Aqua Voice or Willow Voice.

Where VoiceInk is better

Price. $25 once for one Mac, $49 for three, and a 14-day refund window. Speechy Lifetime starts at $99.
Open source. Anyone can read the code, and you can build it yourself for free.
More models. Many local models, plus cloud services with your own API keys.
Modes. Different settings switch on by app, website or spoken word.
An iPhone app, in beta.

The short version

If you want the cheapest good local dictation, or open source matters to you, VoiceInk is a fine choice. Choose Speechy if other voices around you get typed, if you speak more than one language, or if you work with Claude Code and Codex all day. See pricing.

VoiceInk details checked on 14 October 2026 on its pricing page, docs and GitHub. Prices and features change, so check their site before you decide.