Skip to content
Limited-time offer. Save up to 50% on Speechly Premium. See pricing

Text to Speech for Translators: Revise Your Work by Ear

Text to speech for translators: a 2-pass method that catches omissions and calques, plus free setups for Word, Edge, Mac and Trados in any language.

Updated 9 min readBy the Speechly team
A translator listens to his laptop read his page aloud, one finger on the printout and a plum pen ready beside two open dictionaries.
Quick answer

Text to speech for translators works best as a revision check: hear the source while you read the target to catch omissions and mistranslations, then hear the target alone to catch stiff phrasing. Word, Edge and every Mac do it free. Set the voice to the language you're hearing, and keep a visual pass for punctuation and terms.

Key takeaways
  • Use two listening passes: hear the source while your eyes check the target for accuracy, then hear the target alone for fluency.

  • A small 2019 study of translation revisers found better revision quality, especially on accuracy errors, when the source text was read aloud.

  • The voice must match the language it's reading. Word follows the proofing language, Macs use the System speech language, and Windows needs the language's text-to-speech pack installed.

  • Ears miss homophones like French a/à, punctuation conventions, number formats and terminology consistency, so your QA checks and a visual pass still come last.

  • Check your client agreement or NDA before you upload a confidential source or target file to any cloud app.

Text to speech for translators turns revision into a two-sense check: a voice reads the source while your eyes stay on the target, or reads the target on its own, so a skipped clause, a wrong negation or a calque you stopped noticing three passes ago becomes audible. It costs nothing to try. Word, Microsoft Edge and every Mac already have read-aloud voices for dozens of languages.

Below: the two listening passes, how to set the right voice for each language, what works inside a CAT tool like Trados, what your ears will miss, and what to check before a client's file goes into any app. If you want natural voices on every device, Speechly does that on Mac, Windows and your phone.

(Looking for an app that translates text and then speaks the result? That's a translation app: Google Translate, for example, has a Listen button. This guide is for translators checking their own work.)

Why use text to speech for translators?

Because self-revision is hard. You know what the source means and what you meant to write, so your eyes slide over the gap between them. A voice reads only what is on the page, and your brain can't autocorrect what it hears as easily.

There's research behind this. In a 2019 case study in Informatics, Dragoș Ciobanu, Valentina Ragni and Alina Secară at the University of Leeds had 11 participants revise translations with and without speech synthesis reading the source. Revision quality improved with sound, especially for accuracy errors, and most participants preferred listening to the source. It's a small study: a strong hint, not proof.

A 2024 study by a team that included Ciobanu and Secară followed 21 professional translators post-editing English to German machine translation while listening. It found a small gain in quality, a substantial cost in productivity, and a substantial drop in the cognitive effort spent on the target text. The lesson: listening pays off on texts where accuracy matters most, not on every volume job.

How to proofread a translation by listening

The source pass and the target pass catch different errors, so do them separately.

Pass What the voice reads What your eyes do What it catches Speed
1. Accuracy (bilingual) The source, one segment or paragraph at a time Follow the target Omissions, additions, wrong negations, swapped figures, mistranslated terms 0.8× to 1×
2. Fluency (monolingual) The target, start to finish Follow along, or close your eyes for short stretches Calques, source-language word order, clumsy rhythm, sentences too long to follow, dropped or doubled words 1×

Pass 1: listen to the source, read the target

Play one source segment or paragraph, no more, then pause. While it plays, track the target with your eyes and ask one question: did everything I just heard make it across? Missing qualifiers ("not", "only", "up to"), skipped list items and dropped dates stand out when the source comes in through your ears instead of competing with the target for your eyes.

Pass 2: listen to the target alone

Now hear the translation the way its reader will, without the source. This is where source-language interference shows: borrowed word order, literal idioms, a register that's slightly off. If a sentence makes you stop to work out its meaning, a reader will stop too.

You'll also hear the leftovers of your own editing: a doubled word, an article that no longer agrees, half a sentence you forgot to delete. Our guide to proofreading by listening covers this pass in depth, and text to speech for editors applies it to editing.

How to set the right voice for each language

The most important setting is the voice's language. An English voice reading German or French sounds like nonsense and misreads numbers: "1.000" is one thousand in German, but an English voice reads it as a decimal. Set the voice to the language you're listening to before each pass.

Microsoft Word (Windows and Mac)

Word's Read Aloud uses the proofing language set for the document, per Microsoft's Read Aloud guide. It's available in Office 2019, Office 2021 and Microsoft 365.

  1. Select the text, or press Ctrl+A (⌘A on a Mac) for the whole document.
  2. On Windows, choose Review > Language > Set Proofing Language. On a Mac, choose Review > Language. Pick the language and select OK, as Microsoft explains here.
  3. Click where you want to start and choose Review > Read Aloud, or press Ctrl+Alt+Space on Windows or Ctrl+Option+Space on a Mac.
  4. Pause and resume with Ctrl+Space on Windows or Option+Space on a Mac. On Windows, Alt+Left Arrow and Alt+Right Arrow slow down and speed up; on a Mac, use the gear icon in the Read Aloud controls.

Tip for bilingual files: give the source and target their own proofing languages, so each is read by the right voice.

Windows: add a voice for your language

Word, Edge and other Windows apps can only use voices that Windows has installed. Microsoft's voice download guide gives the steps:

  1. Open Settings > Time & Language > Language & region and select Add a language.
  2. Pick your language (languages with voices show a text-to-speech icon) and select Next.
  3. Make sure Text-to-speech is checked, then select Install.
  4. Under Time & Language > Speech, set the Speech language and pick a default voice under Voices.

Not every language has a voice. For a full walkthrough, see how to add more text-to-speech voices in Windows 11.

Microsoft Edge (web pages and PDFs)

Open the page or PDF in Edge and press Ctrl+Shift+U, or right-click and choose Read aloud. In the Read aloud bar, select Voice options to set the reading speed and pick a voice. Microsoft says Edge offers voices from Afrikaans to Zulu, with regional accents such as Canadian, Belgian and Swiss French (Edge Learning Center). The natural-sounding voices need an internet connection.

Mac: Speak selection in any app

  1. Open System Settings > Accessibility > Read & Speak and turn on Speak selection.
  2. Open the System speech language pop-up menu and choose the language you're about to hear.
  3. To add voices, click the Info button next to System voice, choose a language in the sidebar and click a voice to download it (Apple's steps).
  4. Select text in any app and press Option-Esc. Press it again to stop.

Speak selection works in most Mac apps, so you can hear a segment in a CAT tool, a client email or a reference PDF. Switch the System speech language when you move between source and target.

Can you use text to speech in Trados or another CAT tool?

Yes, though built-in speech is still rare in CAT tools. The 2019 Leeds study noted that its authors knew of no CAT or revision tool with speech synthesis at the time. Your options today:

  • Trados Studio: the RWS AppStore lists a Trados TTS app. RWS's documentation says it plays a source or target segment, or the whole file, supports a foot pedal for pausing, and generates the audio with Microsoft's Azure speech service. Check the listing for the Studio versions it currently supports.
  • Any CAT tool on a Mac: select the source or target text in the editor and press Option-Esc for Speak selection.
  • Any CAT tool, any computer: export a target preview or the finished target file as DOCX or PDF, then listen to it in Word, Edge or a reader app. This suits the fluency pass, since it reads the target as one continuous text.

Windows has no simple select-and-speak shortcut like the Mac's. Narrator, the built-in screen reader, can read a selection (Narrator key+Shift+Down Arrow), but it narrates everything you do while it's on. In a Windows CAT tool without speech, the accuracy pass is easier with a plugin like Trados TTS or a reader that works on selected text in any app (see the Speechly section below).

What text to speech won't catch in a translation

Listening is strong on meaning and flow, and weak on anything that doesn't change the sound. Keep your QA checks and a final visual pass for these:

Your ears catch Your ears miss
Omitted or added content Homophones: French a/à, ou/où, ces/ses; English its/it's
Wrong negations and swapped figures, when you compare Punctuation conventions, like French spaces before colons and guillemets
Calques and source-language word order Capitalization, including German nouns
Register and tone that drift Accents and diacritics that don't change the sound
Dropped, doubled and leftover words Number, date and unit formats
Sentences too long to follow Terminology consistency and client glossary choices

Voices have blind spots too. They guess at proper names, acronyms and loanwords, and a source term quoted in the target is often mispronounced. That's the voice, not your spelling. For how voices handle mixed-language text, see can text to speech read other languages.

Is it safe to put a client's files into a text-to-speech app?

Check before you upload. Translation work often falls under an NDA or a client agreement that limits which services can process the text, especially for legal, medical, financial and unreleased product content.

  • Built-in tools like Word's Read Aloud and Speak selection run inside apps you already use for client files. But Edge's natural-sounding voices need an internet connection, so assume online voices send the text to a speech service.
  • Cloud apps and CAT plugins send text to a speech service too. Trados TTS, for example, uses Microsoft Azure. Their documentation and privacy policy should say what happens to it; read them, and ask the client or agency when your agreement is unclear.

Our guide to whether text to speech is private lists what to look for.

How translators can use Speechly

Built-in tools work, but each has its own voices, shortcuts and settings. Speechly is one reader across Mac, Windows, iPhone, iPad and Android, with natural AI voices and word-by-word highlighting.

It has 25+ voices in eight languages: English (American and British accents), Spanish, French, Italian, Portuguese (Brazilian), Japanese, Mandarin Chinese and Hindi. If those cover your language pair, preview the voices and switch between them for source and target passes. For other languages, including European Portuguese, use the built-in voices above.

  • Inside your CAT tool on a Mac: select a segment or paragraph and press ⌘E. A small floating, always-on-top mini player reads it aloud while your eyes stay on the other column.
  • Inside your CAT tool on Windows: select the text and press Alt+R. A floating player reads it aloud, and recent readings are saved.
  • For the fluency pass: open the exported DOCX or PDF in Speechly. Jump to any sentence and start there after a fix, and set the speed anywhere from 0.5× to 2×. See Speechly for PDFs.
  • Away from your desk: listen to a long target on your phone with the screen locked. Speechly saves your place, so you can pick up on any other device you've signed in on.

Content is processed securely and is not shared. Speechly has no offline listening and no audio export, so it's for checking your work by ear, not for producing audio deliverables. The free plan includes two text-to-speech files with premium voices; Premium is $14.99/month or $139/year for unlimited listening on every device.

The bottom line

Text to speech won't replace your QA checks, your glossary or a careful visual read, but it catches the errors that survive silent revision. Hear the source while your eyes check the target, then hear the target alone, with the voice set to the right language each time. Start with the voices already on your computer; when you want natural voices on every device, download Speechly free.

Frequently asked questions

Do professional translators use text to speech?

Some do, mostly for revision and self-revision. Translation researchers have tested it with revisers and with professional post-editors, and Trados Studio users can add the Trados TTS app from the RWS AppStore to hear segments read aloud.

Is there a tool that translates text and reads it aloud?

Yes, that's a translation app rather than a revision tool. Google Translate, for example, has a Listen button that speaks the translated text. It's handy for a quick gist, not for checking a professional translation.

Should I listen to the source text or the translation?

Both, in separate passes. Hearing the source while reading the target helps you catch accuracy problems like omissions, while hearing the target on its own shows you how it sounds to a reader who never sees the source.

Can text to speech read a document with two languages in it?

It can, but most voices read everything in one language. In Word, mark each passage with the right proofing language so Read Aloud switches; elsewhere, expect quoted source terms to be mispronounced and don't treat that as an error.

Does text to speech slow down post-editing?

It can. A 2024 study of 21 professional translators post-editing English to German machine translation found a small positive effect on quality but a substantial negative effect on productivity, so it suits high-stakes texts more than volume work.

Can I use a text-to-speech reader to make audio translations or voiceovers?

Not with a reader app. Readers like Word's Read Aloud or Speechly play text aloud for you; producing audio files for a client needs a speech-synthesis service whose license covers commercial audio.

Can I listen to a translation on my phone?

Yes. Export the target as a Word file or PDF, open it in a reader app, and listen away from your desk, which works well for a final fluency pass on long documents.

Speechly

Listen with Speechly

Natural AI voices on iPhone, iPad, Android, Mac and Windows, with word highlighting and your place saved on every device. Free to download.

Prices and features of other products were checked on November 18, 2026 and can change. Speechly is not affiliated with the other companies named here.