Which languages does it actually support?

Twenty-five, all European: English, French, German, Spanish, Italian, Portuguese, Dutch, Polish, Czech, Slovak, Hungarian, Romanian, Bulgarian, Croatian, Slovenian, Greek, Swedish, Danish, Finnish, Estonian, Latvian, Lithuanian, Maltese, Russian, Ukrainian. The model is NVIDIA's Parakeet TDT v3, which is what the app is named after. Languages outside that list, including Mandarin, Japanese, Korean, Arabic, Hindi, Turkish and Hebrew, are not supported. If the source or target language for your work is in that group, this is not the right tool yet.

Does it really switch language mid-dictation without me changing a setting?

Yes. The 25-language detection runs at the model level on every utterance. You can dictate an English sentence framing a quote, switch to dictating the German source verbatim, switch back to English, all without touching anything. The model identifies the language by itself. Whisper-based dictation tools tend to need an explicit language flag and will mistranslate or anglicize mid-paragraph switches; Parakeet TDT v3 handles them in stride. For translators working with bilingual notes or quoting source material in-line, this is the headline feature.

Does it work with Trados, memoQ, Wordfast, Phrase or Smartling?

Yes. Parakeety pastes the transcript by synthesizing a Cmd+V keystroke into whichever app has focus. That works in Trados Studio, memoQ, Wordfast, OmegaT, plus the browser-based tools Phrase, Smartling, Crowdin, Lokalise, Memsource, plus Word, Pages, Notes and any text field anywhere on macOS. The integration is at the operating-system level rather than at the CAT-tool level, so there is no plugin to install and no segment-aware behavior to configure.

Can I use it for live interpreting?

Not really, and that is a deliberate design choice. Parakeety is push-to-talk dictation for note-taking, not live captioning or simultaneous interpretation. Hold the key, dictate, release; the transcript pastes at the cursor. It is the right shape for an interpreter writing notes between turns or after a session, not for producing a running transcript of someone else's speech. If your need is to caption a live conversation, this is not the right tool.

← Resources

Parakeety for translators and interpreters

Most cloud dictation apps treat language as a setting you change before you start talking. That is awkward for translators, who spend the day moving between a source language they are reading and a target language they are writing, often quoting one in-line inside the other. Parakeety is built differently: local Mac dictation with multilingual auto-detection, running NVIDIA's Parakeet TDT v3 on the Apple Neural Engine across 25 European languages. You hold the section key, dictate a sentence, release; the model figures out which language you spoke in by itself, and the transcript pastes at the cursor in whichever tool you are working in. Switch language mid-paragraph and the dictation follows.

Where dictation fits in translation work

Three workflows where it earns the price. Translators dictating target-language drafts while reading source material on the other side of the screen; the spoken cadence often produces a more natural target-language sentence than the typed one, especially for prose work. Interpreters writing running notes between turns of a session, or tidying up the consecutive-interpretation notes after the session ends, often switching between source and target language in the same note. Bilingual academics and writers quoting source material verbatim into a working draft, switching back to commentary in the working language, without breaking the flow to flip a setting.

In all three the dictation pastes wherever the cursor is. That includes the desktop CAT tools (Trados, memoQ, Wordfast, OmegaT), the browser-based localization platforms (Phrase, Smartling, Crowdin, Lokalise), plus Word, Pages, Notes, or whichever combination of tools you have settled on. Parakeety does not integrate with the CAT tool specifically. The integration is that the operating system already knows where your cursor is, and Parakeety pastes the transcript there, including into the active segment in a CAT-tool grid.

The 25 languages, and the ones it does not handle

The supported list is: English, French, German, Spanish, Italian, Portuguese, Dutch, Polish, Czech, Slovak, Hungarian, Romanian, Bulgarian, Croatian, Slovenian, Greek, Swedish, Danish, Finnish, Estonian, Latvian, Lithuanian, Maltese, Russian, Ukrainian. The auto-detection is what makes it usable for translation work. You can dictate an English paragraph framing a quote, switch to dictating the German source verbatim, switch back to English commentary, all without changing any setting. The model picks up which language you are speaking by itself. Local dictation for writers covers the same multilingual flow from the academic and research-writing side.

Languages outside the supported list, including Mandarin, Japanese, Korean, Arabic, Hindi, Turkish and Hebrew, are not currently supported. If your source or target work sits primarily in those languages, this is not the tool yet. The 25-language scope is a property of Parakeet TDT v3, the underlying model, not of the app wrapping it.

Confidential client work and the architectural answer

A lot of translation work is bound by an NDA. Legal translation, medical translation, contractual material, pre-publication marketing copy, internal comms, court interpretation; the routine for most freelancers is to sign a confidentiality undertaking before the source documents arrive. Cloud dictation tools are awkward in that frame because the audio of you reading or paraphrasing the source material does pass through the vendor's servers, and the transcript is usually retained for some period to support the product. Whether that is consistent with your client's confidentiality undertaking is a question worth asking the client; many will not want to answer it.

Parakeety side-steps it. Audio is captured in memory while the key is held, processed on the Apple Neural Engine, the transcript pastes at the cursor, the buffer is discarded. Nothing is written to disk, nothing leaves the Mac. The only outbound traffic is the one-time speech-model download from Hugging Face on first launch and periodic license checks, which carry only the license key, a hash of a hardware ID, and the machine hostname. For the same architectural framing applied to legal work, see local dictation for lawyers.

Practical fit

Each license covers one Mac. Most freelance translators work on one machine, in which case one license is the model.
Audio is captured to a memory buffer while the key is held. As soon as you release, the buffer goes through the speech model, the transcript pastes at the cursor, and the buffer is discarded. Nothing is written to disk. Nothing is logged.
The model auto-detects across the 25 supported languages, including mid-paragraph switches. There is no language-picker UI and no per-segment setting to remember.
The push-to-talk key is currently the section key (§, below Esc). A user-configurable shortcut is on the roadmap.
First launch downloads the speech model (around 600 MB) and asks for microphone and accessibility access. After that the app lives in the menu bar and runs offline.

What Parakeety is not

Worth being honest. Parakeety is not a translation tool. It does not translate between languages; it transcribes the language you spoke into text in that same language. It is also not a tool for transcribing recorded interviews, lectures or audio files; it is push-to-talk only, and it does not accept audio input. It does not produce parallel-text output, segment-aware paste behavior into CAT tools, or terminology-managed transcription. If your need is for translation, recorded transcription, or CAT-integrated dictation, Parakeety is not the right shape of answer. For dictation as a fast first-draft input across the languages it does cover, with nothing leaving the Mac, it is.

FAQ

Which languages does it actually support?: Twenty-five, all European: English, French, German, Spanish, Italian, Portuguese, Dutch, Polish, Czech, Slovak, Hungarian, Romanian, Bulgarian, Croatian, Slovenian, Greek, Swedish, Danish, Finnish, Estonian, Latvian, Lithuanian, Maltese, Russian, Ukrainian. The model is NVIDIA's Parakeet TDT v3, which is what the app is named after. Languages outside that list, including Mandarin, Japanese, Korean, Arabic, Hindi, Turkish and Hebrew, are not supported. If the source or target language for your work is in that group, this is not the right tool yet.
Does it really switch language mid-dictation without me changing a setting?: Yes. The 25-language detection runs at the model level on every utterance. You can dictate an English sentence framing a quote, switch to dictating the German source verbatim, switch back to English, all without touching anything. The model identifies the language by itself. Whisper-based dictation tools tend to need an explicit language flag and will mistranslate or anglicize mid-paragraph switches; Parakeet TDT v3 handles them in stride. For translators working with bilingual notes or quoting source material in-line, this is the headline feature.
Does it work with Trados, memoQ, Wordfast, Phrase or Smartling?: Yes. Parakeety pastes the transcript by synthesizing a Cmd+V keystroke into whichever app has focus. That works in Trados Studio, memoQ, Wordfast, OmegaT, plus the browser-based tools Phrase, Smartling, Crowdin, Lokalise, Memsource, plus Word, Pages, Notes and any text field anywhere on macOS. The integration is at the operating-system level rather than at the CAT-tool level, so there is no plugin to install and no segment-aware behavior to configure.
Can I use it for live interpreting?: Not really, and that is a deliberate design choice. Parakeety is push-to-talk dictation for note-taking, not live captioning or simultaneous interpretation. Hold the key, dictate, release; the transcript pastes at the cursor. It is the right shape for an interpreter writing notes between turns or after a session, not for producing a running transcript of someone else's speech. If your need is to caption a live conversation, this is not the right tool.

Try it

Parakeety is a Mac menu-bar app. Hold the section key, talk, release; your words paste at the cursor in whichever app you were typing into. Audio never leaves the machine. There is a free 7-day trial with no card required. After that it is $30 once.

Try Parakeety free →