Speak Spanish, type in English

Picture the interpreter who sits in a glass booth at the back of an international meeting. A delegate speaks fast, messy Spanish into the microphone, full of half-finished sentences and filler words, and a half-second later clean, confident English comes through everyone's headphones. That interpreter is doing two jobs at once: translating what was said, and quietly fixing the grammar so the result sounds like something the speaker would be proud to have written. Nobody hears the stumbles. They only hear the polished version.

That is exactly the job a bilingual professional wishes their computer would do. You think in Spanish. Your client is in Miami, Austin, or New York, and the email has to land in English. Right now you are stuck being your own interpreter: composing in your head in one language, translating on the fly, and typing it out while second-guessing every preposition. Your operating system will happily transcribe your voice, but it will not play interpreter. Let us look at why, what you can do for free today, and which tools actually pull off the booth trick.

The gap: your OS speaks one language at a time

Both Windows and Mac ship with free, system-wide voice typing. On Windows it is Voice Typing, opened with Win + H. On Mac it is Dictation. Both are genuinely good at what they do, and both share the same hard wall: they transcribe one language at a time, the one you told the system to expect, and they type that same language back. There is no setting anywhere that means "listen in Spanish, write in English."

The mechanics make this concrete. On Windows, the voice typing language follows your keyboard input language; you change it by switching keyboards with Win + Space, and not every language is even supported for voice input. Microsoft's own Voice Typing support page describes dictating into the selected language and nothing about translating to another one. On Mac, Dictation's language is tied to the input sources you have enabled, and you toggle it by hand before you speak. Apple's guide, Use Dictation on Mac, is the same story: it is dictation, not translation.

So if you set the engine to English and speak Spanish, you get phonetic nonsense. If you set it to Spanish and speak Spanish, you get Spanish, which is not what your client reads. The only native ways to reach English are to force yourself to speak English out loud, which is slower and more error-prone for a non-native speaker, or to dictate Spanish and translate it somewhere else afterward. No native tool does the booth trick in one motion.

The free route: dictate, translate, paste

The honest baseline costs nothing and needs no new app. You dictate in Spanish with your OS, translate the result in a separate window, and paste the English back. It works. It is also three steps every time, and it breaks your flow. Here is the real how-to.

1. 1

   On Windows: dictate in Spanish with Win + H

   Put your cursor in any text box, press Win + H, and speak. If Voice Typing is not already listening in Spanish, switch your keyboard language with Win + Space first. Full instructions are on Microsoft's Voice Typing page.

2. 2

   On Mac: dictate in Spanish with Dictation

   Enable Spanish as an input source, then press the Dictation key (or the shortcut you set) and speak. Apple walks through setup in Use Dictation on Mac. Remember to switch the language back afterward if your default is English.

3. 3

   Translate the Spanish text

   Select the Spanish you just dictated, copy it, and paste it into DeepL or Google Translate. Read the English output once for tone, because a literal translation can sound stiff or subtly wrong to a US client.

4. 4

   Paste the English where it belongs

   Copy the translated English and paste it into your email, doc, Slack message, or form field. Then fix any formatting the round trip flattened, since translators tend to drop line breaks and lists.

That is the free route in full, and for occasional messages it is perfectly fine. Its friction is real, though, and worth naming before you decide it is enough:

• It is at least three steps per message, so it breaks your flow every single time.
• You lose formatting on the round trip: line breaks, bullet lists, and spacing usually flatten.
• You still have to eyeball the translation for tone, because literal is not the same as professional.
• Native dictation mangles proper nouns, so client names, product names, and brands come out wrong and have to be fixed by hand.
• You narrate your own punctuation in the dictation step, saying "coma" and "punto" out loud.

The capability: a multilingual engine plus an LLM pass

Here is the part that makes the booth trick possible in one step. Modern speech models in the Whisper family are multilingual by design; they can transcribe roughly ninety-nine languages and can also translate as they go. Put a language-model cleanup pass on top of that, and the tool can take your spoken Spanish, render it as English, and then smooth the grammar and punctuation so the result reads like the English you meant, not a word-for-word echo of the Spanish you thought. That is the interpreter in the booth, in software.

This is a genuine capability, and it is worth being clear that it is not magic and not unique to any one product. Several tools ship it today. The differences are in how directly they do it, what they cost, and where they run. If your accent tends to trip up dictation in the first place, the transcription quality underneath all of this matters even more; that is its own topic, covered in voice typing that understands your accent.

The honest roundup of tools that do this

Oravo is the most direct match for this exact problem. It markets itself around "speak Spanish, get polished English" (and the reverse), claims support for sixty-plus languages, works system-wide in any text field, and runs an LLM-style cleanup that strips filler and fixes grammar. It has a free tier and a paid Pro plan. Treat its headline accuracy numbers as marketing, not a benchmark, but the core positioning is real and squarely aimed at bilingual writers.

VozFlow is the other purpose-built option. It is aimed specifically at Spanish and English bilingual professionals: one hotkey dictates in your language, and a separate hotkey dictates and translates in the same motion, so you can speak Spanish and get English. It types wherever your cursor is, runs on Windows 10 and 11 and macOS 12 or later, and offers a short free trial with no card. If your whole life is ES and EN, it is a natural fit.

Wispr Flow is a polished paid tool that supports a hundred-plus languages with automatic detection and can even switch languages mid-sentence. Be precise about what that means, though: auto-detect outputs the language you spoke. Its cross-language translation is a separate Command Mode action on the Pro plan, done after the fact, so it is not a default one-step "speak Spanish, type English" mode. Superwhisper also handles a hundred-plus languages and offers optional translation to English plus custom modes people repurpose for translation, but its automatic detection is known to mis-hear non-English speech as English, and the vendor's own docs suggest a vocabulary workaround. Capable, with caveats.

Golem is one more tool built on that same mechanism. It captures at the operating-system level on Windows and Mac, runs OpenAI's Whisper (large-v3-turbo) for transcription with an LLM cleanup pass for punctuation and obvious fixes, and can dictate in one language and output another, so you can speak Spanish and get clean English typed straight into whatever field your cursor is in. It also lets you teach it custom vocabulary so client and product names stop getting mangled. To be honest about the trade-offs: Golem needs a login, and it needs an internet connection for the cleanest results because transcription runs server-side, so it is not an on-device or offline tool. It is free to start with no card, and Pro is $3 a month or $30 a year. It is Windows and Mac only, not Linux and not phone keyboards.

How to choose

Match the tool to how often you actually do this. If you translate a message once a week, the free route is fine and you should not pay for anything. If you write to US clients every day and the language switch is a constant tax on your time, a purpose-built bilingual tool pays for itself in the flow it saves. Weigh three things: whether translation is one step or a bolted-on afterthought, whether it runs on your platform, and whether it can learn your client and product names.

• Occasional translator: use native dictation plus DeepL or Google Translate, free.
• Daily ES to EN writer: try a purpose-built bilingual tool like Oravo or VozFlow, or an OS-level one like Golem, and use the free tiers to compare on your own voice.
• Privacy or offline is non-negotiable: prefer a tool with an on-device model, like Superwhisper's local modes, and accept that server-side tools including Golem are off the table for that use.