How to Transcribe Audio and Video to Text, Step by Step

2026-11-08 ·

Turning a recording into text gives you something you can search, quote and forward. In Magic Master this is audio to text: a recording of a meeting, interview, lecture, podcast or video becomes text with timecodes. It works inside the app: open transcription.

Before you upload: three minutes of preparation

  • Need only a part? Cut it out with the audio cutter: a shorter file uses fewer free minutes.
  • Music under the voice makes recognition harder. For a song or a voice over a beat, split the voice first with the vocal remover.
  • An echoey room swallows word endings. If the recording sounds like a bathroom, remove the echo first.
  • List names and terms: people, companies, jargon. You will need them in step 3.

How it works

  1. Upload a file, audio or video. Only the audio track is taken from video, and the file itself is not kept.
  2. There is no language to choose: Russian speech is recognised.
  3. Add names and terms to "Names and terms": it is a hint for recognition.
  4. Press Transcribe. Long recordings are processed in the background: you can close the tab, we email you when it is ready, and the transcript shows up under "My transcripts".

What you get

  • Text split into paragraphs with timecodes like [12:31].
  • A player next to the text: click a word or a timecode to jump there; the current word is highlighted.
  • Search and editing right on the page.
  • Export: TXT (with or without timecodes), DOCX, SRT and VTT. Copy all the text with one button.

Which format to pick

Format Good for
DOCX Meeting minutes or lecture notes in Word and Google Docs
TXT with timecodes Quotes you want to check against the recording
TXT without timecodes Clean text for an article or a post
SRT or VTT Subtitles for YouTube and video editors

SRT and VTT subtitles

Subtitles are built from the same text with a time for every word: up to 42 characters per line and two lines per cue, each shown for 1 to 7 seconds, with breaks at pauses. If you edit the text, the paragraph keeps its timing and the words inside it are spread again, so the subtitles stay in sync.

How to check the text in five minutes

Automatic recognition handles clear speech well and makes mistakes in names, numbers and places where people talk over each other. Search for the names and figures and listen to those spots by clicking a word. Always listen to a quote in full before you publish it: a missing "not" changes the meaning.

Price

Without an account Signed in
Free 5 minutes a day, text on screen 30 minutes a day, all formats

The limit resets at midnight UTC, and the app shows the minutes you have left before you start. Later, beyond the limit, 1 token per started 30 minutes of recording; the brief, topics and action items are included.

Privacy

The audio is deleted right after transcription, and the text is kept for 30 days. Delete removes a transcript immediately. Recordings are not used to train models. If you transcribe a meeting, get consent from the participants.

Common problems

  • Several people talking at once may leave gaps. Speaker labels are not in the first version.
  • Rare names and terms are best added to the list before you start.
  • Subtitles lag behind the speech: check that you uploaded the same cut of the video that you publish.

A voice message from a messenger? There is a separate guide: voice message to text.

Questions and answers

How much does transcription cost?

Signed in: 30 minutes of recording a day for free. Without an account: 5 minutes, the text is shown on screen, and file downloads need sign-in. The app shows the minutes you have left before you start. Later, beyond the limit, 1 token per started 30 minutes of recording; the brief, topics and action items are included.

Which files work?

Audio (MP3, WAV, M4A, OGG, OPUS including messenger voice messages, FLAC, AAC, AMR, AIFF) and video (MP4, MOV, WEBM, MKV); only the sound is taken from video.

What happens to my recording?

The audio is deleted right after transcription. The text is kept for 30 days, and the Delete button removes it immediately. Recordings are not used to train models. The brief, topics and action items are written by an AI service from the finished text: it receives only the transcript text.

Try Echoteka

Upload a track — the same mastering engine, prices in euros.

Open Echoteka →AI mastering onlineLUFS analyzer