Skip to main content
PromptQuorumBuilt for humans. Structured for AI.
Home/Power Local LLM
Voice, Speech & Multimodal

Univocal Review: Read-Aloud and Dictation on Your Mac

·8 min readBy · Founder of PromptQuorum · Discovery engine for open-weight & open-source AI

Univocal is a Mac app with no published source code that reads selected text aloud and types your dictation, and its maker says both run on your own Apple silicon Mac with nothing sent anywhere. It costs $39 once after a 14-day free trial, with no subscription. The newest release is version 0.3.33 from 11 November 2026. This review rests on the maker's public website, changelog and privacy page; the one first-hand session is in the linked hands-on test.

Key Takeaways

  • Local claim: both directions run on the Mac; the first-run download, licence checks and daily update checks still use the internet.
  • License and maker: proprietary, published by NarNar Labs LLC; no source repository was found and the maker's country is not stated.
  • Price: $39 once with a 14-day trial; the maker also lists time-limited discount codes.
  • Maintenance: version 0.3.33 was released on 11 November 2026, so the project is actively updated.
  • Gaps: Apple silicon and macOS 14 or later only, no named dictation model, and nothing in the pages about code signing.

📍 In One Sentence

Univocal is a paid Mac app that combines text-to-speech and dictation and, per its maker, processes both on your own Mac instead of in a cloud service.

💬 In Plain Terms

Think of two keys on your keyboard: one makes the Mac read what you selected, the other turns what you say into typed text, and the maker says neither sends your voice or text away.

Univocal is a paid Mac app that does two jobs with one install: it reads selected text aloud and it types what you dictate into whatever app has the cursor. Its maker says both run on your own Apple silicon Mac. This review covers what the maker's website, changelog and privacy page document, what they leave out, and how Univocal compares with other voice tools in the directory.

Univocal Review: Read-Aloud and Dictation on Your Mac

Get Univocal

Univocal installs like any Mac app, so no terminal is needed. The official download link serves a disk image, and the speech models are fetched on first launch.

Platform
Get It
MacDownload for Mac — .dmg, Apple silicon, macOS 14 or newer
Free trial (14 days)Purchase page — full product, no card, no account
Setup guideDownloading and first launch
Release notesChangelog
Websiteunivocalapp.com
Privacy policyUnivocal privacy policy
Source codeNo source repository found

The site lists only a Mac build. Windows, iOS and Android appear only as a notify-me form, and Intel Macs are not supported.

Univocal at a Glance

Attribute
Univocal
PlatformmacOS 14+, Apple silicon only
Price$39 once; 14-day trial
LicenseProprietary; no source found
RunsOn your Mac; internet for downloads and checks
Languages32 for dictation and reading (home page)
Real-time transcriptionLive typing as you speak (changelog 0.3.17)
Voice cloningNot stated
Speaker labelsNot stated
Local API serverNot stated; Univocal Connect for local apps

Attributes follow the speech-to-text and text-to-speech comparisons used in the Local LLM Software Directory. A missing statement on the site means unknown, not unsupported.

What Univocal Is

Univocal is a voice utility with two functions: read aloud and dictation. You select text and press a shortcut to hear it, or hold a key, speak and release to have text typed at the cursor, for example in a mail client, a note or a chat.

It is published by NarNar Labs LLC. The maker's pages do not name a founder, a country or a postal address, only a support email. This review rests on the pages listed under Sources; PromptQuorum also ran one first-hand session on a 24 GB M4 Pro Mac (see Related Reading), a first look and not a full evaluation: dictation worked in English, French and German, reading selected text aloud did not start. Voice quality, accuracy and speed are still not rated here.

How to Get Started

Setup is an install, a first-run download and two shortcuts. The steps follow the maker's setup page and changelog.

  1. 1
    Install the app
    Why it matters: Open the downloaded .dmg, move Univocal to Applications and start the 14-day trial; no card or account is needed.
  2. 2
    Let the first launch finish
    Why it matters: The app downloads about 1.2 GB of speech software and models. Version 0.3.11 added the option to start using it before the download completes.
  3. 3
    Grant macOS permissions
    Why it matters: Dictation needs Microphone and Accessibility permission, and reading selected text needs Accessibility permission, according to the maker.
  4. 4
    Read and dictate
    Why it matters: Select text and press Option-Shift-R to hear it, or hold the right Option key and talk. Both defaults can be changed.

Features

The feature set centres on private reading and dictation in any Mac app. Everything below comes from the website, privacy page and changelog.

  • Read aloud. Selected text is read with word-by-word highlighting; with nothing selected, it reads the last copied text. Speed goes up to 4x since version 0.3.16.
  • Dictation. Text lands at the cursor, with punctuation, casing and paragraph breaks inferred as you speak; live typing since version 0.3.17, switched off by default for new installs from version 0.3.28.
  • 32 languages. Reading and dictation share the language list, and the changelog describes mixed-language dictation.
  • Univocal Connect. Other local apps can send text to be read; an Obsidian plugin and a read-aloud extension for Chrome, Edge, Brave, Vivaldi and Arc are listed.
  • History. The last 20 dictations and the last 10 reads are kept on the Mac.
  • Interface languages. English, Korean, Japanese, Simplified Chinese, Spanish, German and French.

Models and Hardware

The maker names Kokoro for reading voices and Apple's SpeechAnalyzer for a live draft on macOS 26, but does not name its dictation model. The privacy page says speech software comes from PyPI and models from Hugging Face, about 1.2 GB in total, and that a local engine listens on the Mac at 127.0.0.1:8787. Whisper and Parakeet are not mentioned on any page read.

Requirements are an Apple silicon Mac and macOS 14 or later. Some reading languages, Korean and German among them, fetch about 400 MB more on first use. No RAM figure and no speed or accuracy numbers are published.

Privacy, License and Cost

The privacy page says nothing you dictate or read is sent anywhere, and that the app has no account, no analytics and no crash reporting. The internet is still involved elsewhere: downloads from PyPI, Hugging Face and, if the Mac has no Python, GitHub; a licence check with Lemon Squeezy that sends the licence key and the Mac's name; and a daily update check against the maker's site. Issue reports are sent only when you press Send Report. The website, unlike the app, uses Cloudflare Web Analytics. The code is not published and PromptQuorum did not examine network traffic, so these statements are the maker's own.

It is a paid, proprietary product: $39 once, one licence for up to three Macs, a 14-day trial and a 30-day refund. The purchase page also lists a launch code and a Black Friday code, both valid until 30 November 2026; check the page for the price at the time you buy. The maker's comparisons with other apps on its own site are the maker's claims and were not verified here.

Trade-Offs: Benefits vs. Limitations

Reading and dictation in one app

What it means in real use:
One install, one licence and one history instead of two tools.
Limitation / caveat:
Mac only; no Intel, Windows or mobile version.

Processing on your Mac

What it means in real use:
Speech and text stay on the machine, per the maker.
Limitation / caveat:
Closed code, so the claim cannot be checked; first launch needs about 1.2 GB of downloads.

One-time price

What it means in real use:
$39 once, up to three Macs, no subscription.
Limitation / caveat:
Discount codes run until 30 November 2026, after which the home page lists $39.

32 languages

What it means in real use:
Read and dictate in many languages, including mixed-language dictation.
Limitation / caveat:
No per-language quality data; the dictation model is not named.

Connections to other apps

What it means in real use:
Obsidian and a browser extension can send text to be read.
Limitation / caveat:
According to the privacy page, the extension runs on all web pages; read its permissions first.

Frequent releases

What it means in real use:
Version 0.3.33 shipped on 11 November 2026.
Limitation / caveat:
Version 0.x; the earliest changelog entry is 0.3.0 from 29 July 2026, so the app looks young.

Univocal vs. Alternatives

These are the closest voice tools in the directory: a free read-aloud app, a free studio that does both jobs, and three Mac dictation apps.

Univocal

Platforms:
Mac
Price and license:
$39 once, proprietary
Role:
Read aloud and dictation
Key difference:
Both jobs in one paid Mac app
Also mentioned in Univocal (1)

Also mentioned in:

Parrot

Platforms:
Mac, Windows, Linux
Price and license:
Free, MIT
Role:
Read-aloud only
Key difference:
Free and open source; no dictation
Articles about Parrot (1)

Voicebox

Platforms:
Mac, Windows, Linux
Price and license:
Free, MIT
Role:
Voice studio with dictation
Key difference:
Adds voice cloning and an MCP server
Articles about Voicebox (1)

Weesper Neon Flow

Platforms:
Mac, Windows
Price and license:
Trial, then 5 EUR/month or 99 EUR once
Role:
Push-to-talk dictation
Key difference:
Whisper on the device; adds Windows, no reading
Articles about Weesper Neon Flow (1)

Also mentioned in:

Sprich

Platforms:
Mac
Price and license:
Freemium, proprietary
Role:
Mac dictation
Key difference:
German-first with a free plan; no reading
Articles about Sprich (1)

Also mentioned in:

MacWhisper

Platforms:
Mac
Price and license:
Freemium, proprietary
Role:
Mac transcription app
Key difference:
Built for files and meetings first
Articles about MacWhisper (2)

Also mentioned in:

Platform, price, and feature details for third-party apps change frequently. Verify current specifics on each project's own page before deciding.

Compare in directory

Who Should Use Univocal

  • Mac users who both listen and dictate. Reading and dictation share one app and one licence.
  • People who avoid cloud voice services. Both functions run on the Mac, per the maker.
  • Multilingual users. 32 languages for reading and dictation.
  • Buyers who prefer paying once. No subscription is offered.

Who Should Not Use Univocal

  • Intel Mac, Windows, iPhone and Android users. Only an Apple silicon Mac build exists.
  • Anyone who needs auditable software. No source code was found and the privacy claims cannot be inspected.
  • Anyone who wants free software. After the trial it is paid; Parrot is free for reading and Voicebox is free for both.
  • Anyone who must work with no internet at all. The first launch needs about 1.2 GB of downloads.

Frequently Asked Questions

Does Univocal work offline?

The purchase page says dictation works offline after the first-run download. The privacy page describes the licence check as failing open, so it should not block use when offline. PromptQuorum did not test it offline.

Is my voice sent to the cloud?

The privacy page says nothing you dictate or read is sent anywhere. On macOS 26, live typing also uses Apple's on-device speech recognition. The code is not published and PromptQuorum did not inspect network traffic.

Which speech model does Univocal use?

The pages name Kokoro voices for reading and Apple voices on older macOS versions, but not the dictation model. Whisper and Parakeet are not mentioned.

Can I use one licence on several Macs?

Yes, up to three; a Mac can be moved off the licence in Settings, according to the purchase page.

Who makes Univocal?

NarNar Labs LLC. The pages do not state its country or a founder.

Verdict

Univocal is a good fit if you have an Apple silicon Mac, want text-to-speech and dictation in one app where, per the maker, everything runs on your own computer, and accept paying once for software whose source code is not published. It has a 14-day trial, a one-time price and a very recent release.

The caveats are scope and verifiability: no Intel, Windows or mobile version, no published code, an unnamed dictation model, and privacy claims that PromptQuorum could not check. Use the trial before you buy.

Sources

← Back to Power Local LLM

MCP