Business Email Translator

Business email fails in translation at the salutation, the request and the sign-off — the three places where conventions differ most between languages and where a literal rendering sounds abrupt or over-formal. The local pass substitutes those conventional phrases and tells you how many it handled.

The Parts of an Email That Are Convention, Not Meaning

An email has a body that carries information and a frame that carries etiquette. The frame is where translation goes wrong, because the conventions are arbitrary: the phrase that reads as politely indirect in one language reads as evasive in another, and a literal rendering of an English request can land as a command. A phrase table is a reasonable instrument for this specific problem, because the frame is made of a small number of recurring stock phrases.

• A compiled table of around a hundred and fifty conventional phrases is grouped by target language, covering the openings that establish the reason for writing, the formulas for making a request, the phrases that soften a deadline, and the closings that precede a signature.
• Substitution runs longest phrase first so that a multi-word formula is replaced as a unit. Performing the same replacements in the opposite order would break longer expressions into partially translated fragments, which is the failure mode that makes machine-translated correspondence obvious.
• Six target languages are supported: Chinese, Spanish, French, German, Portuguese and Japanese. Each has its own phrase set rather than a shared list, since the conventional formulas are not translations of each other.
• A two-position register control offers Formal and Casual. In practice this is the setting that decides whether the request formulas come out in the form used for a first contact or the form used between colleagues who know each other.
• The count of substituted phrases is reported after each run, which is the honest indication of how much of the email was actually handled — a long email with three substitutions is largely still in the source language.
• The result can be copied in one action as plain text, so it drops into a mail client without carrying formatting.

Getting the Frame Right Before the Body

1. Choose the target language first. The phrase sets are separate, and switching language after generating regenerates rather than translating the previous output.
2. Set the register. Formal for a first contact, an external client or anything where the relationship is asymmetric; casual for a recipient you already correspond with weekly. This single setting has more visible effect on the outcome than any other control on the page.
3. Paste the email including the greeting and the sign-off. Both are pure convention and both are in the phrase table, so an email pasted without its opening line is being processed on the wrong text.
4. Read the substitution count, then read the output. If the count is low relative to the length, most of what you are looking at is unmodified source text, and the sensible next step is to have the body translated properly and keep this pass for the frame.
5. Check the two places convention matters most in the output rather than reading top to bottom: the request and the closing. Those are the lines a recipient judges the tone by, and they are the lines the table covers.
6. Keep the original alongside the result for anything consequential. Substitution returns a text of the same shape as the input, so differences are easy to miss without comparing.

One Request in Two Registers

Take a short follow-up email that asks a client to confirm a date and closes politely, and run it with the register set each way.

• With the register set to formal, the request formula is substituted into the form used for an external party and the closing takes the standard courtesy phrase. The sentence that carried the deadline remains structurally as written, because a stated date is content rather than convention and the table does not touch it.
• Switching to casual changes the same two positions and nothing else. The deadline sentence is byte-for-byte identical between the two outputs, which is the clearest demonstration that the register control affects convention rather than the substance of the message.
• The greeting line is substituted in both runs, and this is the change most worth checking against your own knowledge of the recipient. A greeting that is correct for one organisation reads as distant in another, and no phrase table can know which you are writing to.
• If the body of the email contains no conventional formulas — a message that is entirely a paragraph of project detail, for instance — the substitution count comes back low and the output is close to the input. That is an honest result rather than a failure, and the tool reports the count so you can see it.
• Pasting the same email in without its sign-off drops the closing substitution, because there is no closing line to match. The count falls by one and the output ends abruptly, which is the quickest way to confirm the table is operating on phrases rather than on structure.

What the Local Pass Handles and What It Does Not

Alternative	Notes
Conventional frame, local pass	Greetings, request formulas, softening phrases, closings. These are stock expressions and a phrase table handles them accurately and instantly, with the additional benefit that nothing leaves the browser. This is the half a table does better than a model, because the correct rendering is convention rather than meaning.
Informational body, local pass	Not handled. Sentences carrying project detail, dates, figures and commitments are left as written. Reporting the substitution count is what makes this visible rather than leaving you to assume the whole email was processed.
Informational body, connected provider	Rewritten into the target language with the register you selected. This is the step that requires sending the email text to a provider, and business correspondence often contains names, figures and commitments that make that a real decision rather than a formality.
A human translator	Still the right answer where the relationship matters more than the time saved — a negotiation, a formal apology, a first approach to a major client. The frame can be automated; the judgement about how a specific recipient will read a specific sentence cannot.

How the Substitution Works

The email is scanned for phrases from the target language's table, longest first, and each match is replaced in place. Matching is case-insensitive so a formula capitalised at the start of a sentence still matches, and the replacement is inserted as written in the table rather than being re-cased, which is deliberate — the conventional form of a greeting in many languages does not follow English capitalisation rules. Because the operation is a replacement rather than a synthesis, it is reversible in practice: comparing the input and the output tells you precisely which phrases changed and nothing else did.

• No account, no key and no network request are involved in the local pass. The phrase tables are part of the page.
• The two-position register selects between two sets of formulas per language rather than adjusting the output's tone in any graded way.
• Nothing is stored between sessions, so there is no history of previously translated correspondence to leak or to clean up.
• If you connect a provider, the email text is transmitted. For correspondence containing commercial terms that is a decision to take deliberately, and the local pass is the option that avoids it.

Questions About Translating Correspondence

Will the recipient be able to tell this was processed by a tool?

They can usually tell it was not written by a native speaker, and the tell is almost never the frame. Greetings and closings come from a table of conventional formulas and read correctly. What gives it away is the body, which in the local pass is unchanged and therefore entirely in the source language, or in the AI path is fluent but occasionally chooses a register that a native speaker would not use with that particular recipient.

Does the Formal or Casual setting change the whole email?

No. It selects a different set of conventional formulas for the same positions — the request, the softening phrase and the closing. Sentences carrying content are unaffected, and you can confirm this by generating both versions and comparing: the deadline sentence and the project detail will be identical between the two.

Why was most of my email left unchanged?

Because the local pass substitutes recognised conventional phrases and the counts reported afterwards tell you how few there were. An email that is mostly project detail contains little conventional language, so little is substituted. The tool reports the number precisely so that this does not go unnoticed; treating an unchanged body as a translation would be the mistake.

Is there a version of this that keeps everything on my machine?

The local pass is that version. The phrase tables are compiled into the page, the substitution runs there, and no request carries the email — verifiable in the network panel. It handles the frame and not the body, which for a short piece of correspondence is often most of what actually needed translating, since the informational part is frequently names and dates that do not translate.

Last verified: October 3, 2026

Launch Business Email Translator

All Tools · AI Tools · About · How We Test · Contact · Privacy

Maintained by Huang Bin · Last updated October 3, 2026