Text Transcription: turn a recording into text

On the Tools tab the tool is called Text Transcription, and the screen it opens is titled Speech to Text. It suits voice messages, interviews, lectures, podcasts and any other recording where you need the words as text.

What you can add

Text Transcription works with sound. A video is accepted too: the app takes the sound out of it on your device and uploads only the sound.

How to get a transcript

1. Open the Tools tab and tap Text Transcription.

2. Add the audio: Choose for a file, Record audio to record, or Paste link.

3. Wait. Recognition starts by itself as soon as the file is added; there is no Transcribe button. The text appears on the Text tab.

4. Open the Essence tab for the short version. It is written automatically after the transcript.

The spoken language is detected automatically. You cannot pick it by hand.

What you get on the Text and Essence tabs

The Text tab holds the full transcript. With the default recognition engine, when several people talk, it is split into speakers (Speaker 1, Speaker 2 and so on) with a timecode for each part. Tap a timecode to play the recording from that moment. Turn off Show timecodes to see plain text. The text can be copied or shared.

The Essence tab is a short summary written in the language of the app, whatever language the recording is in. It has three parts:

• Brief: a title, a summary, quotes and the main takeaway.

• Key points: the main points with timecodes. Depending on the recording you also get Steps, Questions and answers, Terms or Action items.

• Chapters: the recording split into parts with timecodes.

Export to PDF saves the Essence as a document. Under the transcript, the What next? block offers Discuss in chat, Translate text, Text to speech and Analyze text.

Every result goes to the tool's History. Without Premium the last 3 entries are kept, see how history works.

How to change the recognition engine

Text Transcription has no settings of its own. It uses the engine chosen in Audio Translation: open Audio Translation, tap the settings icon and go to the Recognition tab.

The choice applies to both tools.

How much it costs

With ElevenLabs Scribe, recognition costs about 0.73 tokens per minute of sound, so an hour is about 44 tokens. The Essence is charged separately: a fraction of a token for a short voice message, a few tokens for a long recording. Exact rates for every engine are in token prices for each tool.

What Text Transcription does not do

• It has no field for typing or pasting text. To work with a text you already have, use Text Translation or the chat.

• It does not make subtitle files (SRT, VTT). Burned-in subtitles on a video are made by Add Captions.

• It does not translate by itself. Use Translate text under the transcript, or Audio Translation to get a voiced translation.

Frequently asked questions