What Is Automatic Speech Recognition (ASR)? Definition

What is automatic speech recognition? Often shortened to ASR, it is the technology that identifies the words in spoken audio automatically, without a human typing them. The output is plain text, sometimes with timestamps. Accuracy depends heavily on recording quality, accents and the language spoken.

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

Modern ASR systems are neural networks trained on large collections of recordings paired with their transcripts.
ASR is also called speech-to-text; both names describe the same step of turning sound into written words.
Systems are compared with word error rate, a count of wrong, missing and extra words against a reference text.
Extra steps such as punctuation, speaker labels and number formatting are often added after the recognition itself.
The same engine can behave very differently on a studio microphone and on a phone held across a noisy table.

What it looks like

audio  ->  features  ->  acoustic + language model  ->  text

example result:
{
  "text": "the meeting starts at nine",
  "segments": [
    {"start": 0.0, "end": 2.4, "text": "the meeting starts at nine"}
  ]
}

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

Frequently asked questions

Is ASR the same as speech-to-text?

Yes, in everyday use. ASR is the technical name for the process, and speech-to-text is the plainer name for the same task and its result.

Does ASR understand meaning?

Not in a human sense. It predicts the most likely words for the sounds it hears, so it can write a wrong word that sounds right.

What makes ASR results worse?

Background noise, overlapping voices, strong accents, poor microphones and rare names all reduce accuracy. A clean recording is the biggest single help.

How is ASR quality measured?

Mostly with word error rate, and with character error rate for languages written without spaces. Both compare the output with a human reference text.

How do I use ASR on a recording?

Send the file to the Potok bot on Telegram. Voice messages, audio and video files up to 2 GB are accepted, and the text comes back as a message or file.

Sources

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

Speech recognition (Wikipedia), Wikipedia · checked 2026-10-18
Word error rate (Wikipedia), Wikipedia · checked 2026-10-18

Try it on your own recording