Skip to main content
Accessibility

Speech to textAutomatic speech recognition (ASR)

Speech to text (STT), also called automatic speech recognition (ASR), is the technology that converts speech into text. An AI model analyses the sound of a recording and writes what is said, word by word, with the moment each word is spoken. It is what produces automatic transcripts and subtitles.

How speech to text works

Processing happens in three steps, in a few minutes for an hour of recording.

  • The model finds the passages of speech and identifies the spoken language
  • It converts the sound into words, then adds punctuation and capitals
  • It timestamps each word and, depending on the tool, separates the speakers

What speech to text gets wrong

It writes what it hears, not what was meant. Proper nouns, acronyms and specialist vocabulary are its usual errors, along with passages where several people speak at once. A glossary of terms and a human review are still needed before publishing.

Speech to text in Videas

Videas Scribe runs speech to text on your audio and video in more than 100 languages, then opens the result in an editor: you correct the text, feed a glossary with the terms that keep coming back, and publish the subtitles in the player or export the transcript.

Frequently asked questions

What does speech to text mean?+

It is the automatic conversion of speech into text by an AI model. It is abbreviated STT.

What is the difference between speech to text and ASR?+

None in practice. ASR, for automatic speech recognition, is the technical term; speech to text describes the result.

How accurate is speech to text?+

On clean audio with a single speaker, more than 95% of the words are right. Noise, strong accents and jargon pull that figure down, hence the review.

What is the difference between speech to text and text to speech?+

They run in opposite directions: speech to text turns speech into text, text to speech turns text into a voice.

What is speech to text used for in video?+

To produce a video's subtitles and transcript without typing them, then translate them, and to make the video's content readable by a search engine.