Spanglish to English: how to turn mixed speech into polished text

If you grew up between Spanish and English, you probably don’t think in one language at a time. A sentence starts in Spanish, lands a key word in English, and switches back before the period. Out loud, with friends or family, that mix is effortless — everyone understands. The problem starts the moment you have to type it into an email, a support ticket, a Slack message, or a form that expects clean, professional English.

So you slow down. You translate in your head. You second-guess whether “mándale un follow-up” should be “send him a follow-up” or “follow up with him,” and by the time you’ve rewritten the thought three times, the ten-second message has cost you two minutes. This post is about why that friction exists and how to remove it.

Spanglish isn’t a translation problem

The instinct is to reach for a translator. But translation assumes you’re moving from Language A to Language B — one clean source, one clean target. Spanglish has no single source language. A line like “diles que la factura ya está pagada, les mando el receipt ahorita” is already half English. Feed it to a translator and it either mangles the English words, “corrects” your intent into something generic, or trips over the switch mid-sentence.

What you actually have is an intent expressed in a mix and a target register the situation demands — polite-but-firm, client-ready, clinical, casual. That’s not translation. That’s composition.

Dictation doesn’t help either

Standard voice-to-text was built for one language and clean speech. Speak Spanglish into it and one of two things happens: it forces every sound into Spanish spelling, or it forces everything into English and loses the words it can’t place. Either way you get a verbatim transcript — filler words, spoken grammar, false starts and all — in the wrong register. You still have to edit heavily, which defeats the point of speaking in the first place.

What “Spanglish to English” should actually mean

The goal isn’t a transcript of what you said. It’s the text you meant to write. That takes two things working in sequence:

• Understand the blend. Read the intent inside code-switched speech — Spanish, English, and the switches — instead of forcing it into one language.
• Compose for the moment. Use what’s around your cursor — the email you’re in, the thread, the register the situation calls for — to write finished text, not a raw dump of words.

Do only the first and you get a correct but generic transcript. Do only the second and you get an assistant that can’t hear how you actually talk. You need both. That’s the difference between thirty seconds of speech and text you can send without touching.

A concrete before-and-after

Say you put your cursor in a reply to a client and speak, naturally: “mándale un follow-up a Carlos sobre la invoice, polite pero firme”. You didn’t dictate a sentence — you described an intent in a mix. Finished English, in the right register, reads more like:

Hi Carlos — just following up on the invoice, which is now past its due date. Could you confirm when payment will be processed? Happy to resend it if that helps.

You didn’t translate anything. You didn’t edit a transcript. You said it the way it came out, and the finished version landed at your cursor.

How Brocaly does it

Brocaly is built for exactly this. Put your cursor anywhere — an email, a form, a chat, a doc — speak in your natural Spanish-English mix, and it writes register-appropriate English right there. It reads the surrounding context so the output fits the moment, and it can pull details from your own history when you ask. It’s not dictation and it’s not translation — it answers the question you actually have: what should this text be?

If you spend your day switching between Spanish and English and typing in the second one, that gap is worth closing. See how voice-to-text, dictation, and translation differ or read up on why keyboards fail at code-switching.