Transcribe Multiple Languages Simultaneously

Today we launch one of the most requested features among our enterprise users: Transcription of conversations where multiple different languages are spoken.

Product
Mattias Fält
September 5, 2024
5 min read
Back to blog
Transcribe Multiple Languages Simultaneously

Today we launch one of the most requested features among our enterprise users: Transcription of conversations where multiple different languages are spoken.

When are multiple languages needed?

It's more common than you might think for a conversation to take place in different languages. Several of our customers regularly need to transcribe conversations with multiple different languages, for example:

  • Interviews where the interviewee is more comfortable speaking their native language.
  • Meetings where languages that everyone understands are mixed, e.g., Swedish and English.
  • Interpretation calls where translation occurs between two or more people. In these cases, it's often a requirement that the transcription is in the language being spoken. It's important to be able to go back and verify what has been said and you don't want to rely on a translated text.

Why is a new solution needed?

Previous solutions for transcribing multiple languages simultaneously are often time-consuming, poor, or give incorrect results. We have looked at others' solutions, which often build on the entire text first being transcribed in one language. This often leads to a series of different problems:

  • The language used is inconsistent and doesn't match what is spoken.
  • The audio needs to be transcribed multiple times in different languages and you then have to manually piece together the different parts.
  • The entire transcription becomes in the most common or first language spoken.
  • Sentences are transcribed in one language and then translated back to the spoken language.

We don't think these alternatives are good enough, so we have developed our own solution!

What Klang.ai does

We have developed a completely new method for handling multiple languages in the same conversation. We first identify which languages are spoken before the audio is transcribed. Then we can transcribe each part of the conversation in the right language directly, without having to rely on translations of the text. Each part of the conversation can also be transcribed by the model that is best at that particular language, which guarantees that the transcription maintains the highest possible quality!

From today, this is available for selected enterprise customers.

Are you also interested in transcription of files with multiple languages? Get in touch and we'll tell you more!