Question · Languages and accents
How do you talk to text in Spanish and get English?
Talk to text writes the language you speak, so it does not turn Spanish speech into English by itself. Dictate in Spanish, check the Spanish, and then translate it, or send a finished recording to a Whisper translation model that outputs English.
Updated October 17, 20264 min read7 cited facts, each traced to its source
Talk to text does not translate. It writes the language you select, so if you speak Spanish with Spanish selected, you get Spanish text. To get English from Spanish speech, you need one of two routes: dictate in Spanish and then translate the text, or send a finished recording to a speech model that translates into English. The first route works with tools most people already have. The second is for recordings and needs a technical setup.
Dictation writes the language you select, so English needs a second step
Dictation on a Mac, an iPhone or a Windows PC writes in the language you have selected. On a Mac, you switch between dictation languages by clicking the language label next to the cursor or by pressing the Globe key and choosing a language 1. Apple describes this as switching between languages 2, and the Apple page we read does not describe recognizing two languages within one sentence.
The iPhone keyboard can pair English (U.S.) with Spanish (Latin America) for some language pairs 3. That lets you switch between the two languages without changing keyboards. Apple's description does not mention translation, so do not expect the English to appear on its own.
Route one: dictate in Spanish, check it, then translate
This route is two separate tools, and it works on any device that can dictate Spanish.
- Dictate in Spanish, with the Spanish variant you speak selected. The Spanish speech to text guide explains how to choose the variant.
- Read the Spanish back and fix the errors before you translate. Missing accents, wrong names and missing question marks all reach the translation, and a translator may not fix them for you. The accent mark question covers the checks.
- Paste the Spanish into a translator, and read the English next to your Spanish. Check numbers, names and any sentence that carries a warning or a date.
The two steps add up. An error in the dictation can turn into a wrong English sentence, and a translator can also make its own mistakes, so the English needs its own reading.
Route two: a Whisper model can output English from a recording
OpenAI's speech to text guide names whisper-1 as the model to use for translating a completed audio recording into English 4. This route takes a finished recording, such as a voice note or a meeting clip, and returns English text. It is not live dictation into an app, and it needs a developer or an app that calls the model through an API.
Choose the model with care. The Whisper README, published by OpenAI, says the turbo model is not trained for translation tasks, so choose another model when you translate 5. Whisper supports 98 languages 6, and Spanish is one of them 7.
| Route | What you do | What you get | Main caveat |
|---|---|---|---|
| Dictate, then translate | Dictate Spanish, check it, paste into a translator | English text from a Spanish draft | Errors from both steps add up |
| Whisper translation | Send a finished Spanish recording to a translation model | English text from the audio | Needs an API setup, not live dictation |
| Bilingual keyboard | Switch between English and Spanish while you type | Text in each language you type | Does not translate |
English to Spanish works the same way in reverse
The dictate-then-translate route works from English into Spanish too: dictate in English, then translate the text. The recording route does not help here, because the Whisper translation route outputs English 4. Read the Spanish you get back for accent marks and question marks before you send it, because the translation step can change the sentences you dictated. Treat the Spanish as a draft until you have read it.
It depends on the length, the mix and the recording
- Short, clear sentences are the easiest to check, because each error stays within one sentence. Long dictations give the translator more chances to change the meaning.
- Mixed Spanish and English in one sentence is harder. The Apple page we read describes switching between dictation languages, not mixing two languages in one sentence, so a sentence that mixes both may come out mostly in one language. The code-switching entry explains the problem, and the multilingual dictation guide covers the options.
- Names, places and slang are where the recognizer and the translator are most likely to fail. Check them against your original.
- Recordings with noise make the recognition step weaker, so the translation starts from a less accurate Spanish text.
What to do
- Choose route one for dictation and route two for a finished recording.
- Dictate in Spanish with the correct variant selected, and keep sentences short.
- Read the Spanish once for accents, names and question marks, then fix them.
- Translate, and compare the English with your Spanish sentence by sentence.
- If the English changes the meaning, dictate the sentence again in shorter parts and translate each one.
For the wider picture, see the Spanish speech to text guide, the speech to text translation guide and the question on whether Spanish speech to text is as accurate as English. The speech translation entry defines the terms used here.
Evidence
Every number on this page, with where it comes from. Open a fact to see the exact words in its source.
- 1.
- 2.
- 3.
- 4.
- 5.
- 6.
- 7.
Go deeper
Related questions
- Does Spanish dictation add accent marks automatically?
- Is speech to text as accurate in Spanish as in English?
- Which Spanish should you choose for dictation?
- Can speech to text handle Hinglish, Spanglish and other mixed languages?
- Can Whisper translate audio?
- Can you dictate in two languages at the same time?
- Can you train dictation software to your accent?
- Can you translate Zulu or Xhosa audio to English?