What Is Language Identification in Speech?

What is language identification? It means detecting which language is spoken in a recording, usually before the words are recognised. A detector listens to short stretches of sound and picks the most likely language from a fixed list. Its choice then decides which recogniser reads the speech.

Transcribe a recording

Send your voice message, audio or video to @potok_now_bot on Telegram and the text comes back in the chat.

Open @potok_now_bot in Telegram

Files up to 2 GB · export to TXT, MD, DOCX, PDF, SRT and VTT · speaker labels and timestamps · about 30 languages

Key facts

A detector chooses only from the languages it was built for; speech outside that list is matched to the nearest one and the guess is wrong.
Short clips give the detector little to work with, so one spoken sentence is far less reliable than a few minutes of talk.
Mixed speech is a hard case: a speaker who switches language mid-sentence may be labelled with one language for the whole file.
Related languages with shared vocabulary, such as Spanish and Portuguese, are confused more often than languages that sound very different.
Music and background noise also push the guess off, because the detector hears a signal that is not clean speech.

What it looks like

// illustrative output of a language detector for one recording
{
  "file": "call.m4a",
  "detected": "es",
  "scores": { "es": 0.91, "pt": 0.07, "it": 0.02 }
}
// a low top score points to a short, noisy or mixed clip

See the full list of languages, the tools and the glossary.

Frequently asked questions

Why does the language of a recording matter?

A recogniser expects one language at a time. If it is told the wrong one, it still produces words, but they are the closest sounds in that language and make little sense.

Is language identification the same as speech recognition?

No. Identification only names the language. Recognition turns the speech into words. Identification often runs first so the right recogniser can be chosen.

What happens with a recording in two languages?

Results vary. Many systems pick the dominant language and misread the rest. Splitting the file at the point of change and handling each part separately is more dependable.

How many languages does the Potok bot handle well?

No. The bot works well for about 30 languages with good published results, and a few more with a warning. Check the language list before sending a rarer language.

How can I help the detector?

Trim long silence and music from the start, use a clear recording and keep one language per file where you can. A longer sample of speech helps it decide.

Sources

Each source was opened by hand on the date shown. Pages change, so open the source if a detail matters to you.

Language identification (Wikipedia), Wikipedia · checked 2026-10-18
Speech recognition (Wikipedia), Wikipedia · checked 2026-10-18

Try it on your own recording