Can one pass handle two languages in the same file?

It depends on the system. Many recognisers assume a single language per job, so heavily mixed audio may need segmenting or a model that explicitly supports code-switching.

Should English loanwords be translated?

Usually not in a verbatim transcript, since the speaker chose them. In a localized deliverable it becomes an editorial decision.

How are multilingual transcripts formatted?

Most style guides ask you to keep each passage in its own script, and some ask for a language tag on passages in the secondary language. Confirm with the client.

Should each language be in its own transcript?

For subtitle delivery, yes, because one file carries one language. For a research or record transcript, keeping the languages interleaved as spoken is usually what is wanted.

Definition

What Is Multilingual Transcription?

VideoText editorial team. Updated September 13, 2026.

Direct definition

Multilingual transcription produces text for speech in more than one language — either one file per language, a language-id plus recognition pipeline, or a single transcript that follows code-switching.

The sections below explain how multilingual transcription is used in transcription and subtitle work, what people often mix up, and which VideoText workflow applies when the next step is a file or review pass.

Key takeaways

Recognizing Spanish audio is not the same as translating an English SRT into Spanish.
Code-switching inside one conversation is harder than a single-language file.
Recognizing speech in another language is not the same job as translating existing subtitle cues.

Multilingual transcription vs subtitle translation

Transcription starts from audio. Translation starts from existing text cues and should preserve timecodes.

Code-switching within one recording

Many real recordings move between languages mid-sentence, not just between speakers. A bilingual interview may carry English technical terms inside Hindi or Spanish grammar, and a faithful transcript keeps both.

This is harder than transcribing either language alone. A recogniser set to one language will attempt to force the other into it, producing plausible-looking nonsense rather than an obvious error.

Deciding the deliverable before starting

Three different outputs are commonly requested and often confused: a transcript in the languages as spoken, a transcript entirely in one language with the other translated inline, and a transliteration of the non-Latin passages into Latin script.

Agree which of the three is wanted, and how mixed-script passages should be formatted, before the work starts. Reformatting afterwards is far more expensive than deciding at the brief stage.

Frequently asked questions

Can one pass handle two languages in the same file?: It depends on the system. Many recognisers assume a single language per job, so heavily mixed audio may need segmenting or a model that explicitly supports code-switching.
Should English loanwords be translated?: Usually not in a verbatim transcript, since the speaker chose them. In a localized deliverable it becomes an editorial decision.
How are multilingual transcripts formatted?: Most style guides ask you to keep each passage in its own script, and some ask for a language tag on passages in the secondary language. Confirm with the client.
Should each language be in its own transcript?: For subtitle delivery, yes, because one file carries one language. For a research or record transcript, keeping the languages interleaved as spoken is usually what is wanted.

Related terms

Sources

Mozilla: Common Voice multilingual dataset