Edit while transcribing: Have your story ready when the statement ends

How text can flow and be edited at once

The text on screen has two states:

• Faint text is the engine’s current guess. It may change as new audio comes in, and you cannot edit it.
• Solid text is final and yours. Click it, type, delete or make it bold.

The engine only adds new words to the end of the paragraph it is writing, so a correction higher up stays intact. Undo also only takes back your own changes, not what the engine wrote. While you work higher up, the text keeps flowing below, and the Back to live button takes you to the last line.

AI transcription accuracy explains why faint text comes first.

A press statement, minute by minute

During an invented ministry statement, a reporter’s screen might go like this:

This table is not a promise, as it varies with your device and the stream. The text runs a little behind the speech, and the transcript page shows by how much. For live text we recommend a computer with a graphics card. Steps by type of broadcast are on the live stream transcription and live TV transcription pages.

Tell speakers apart, name them, find them

With Tell speakers apart on, a new paragraph starts when the voice changes, and a label such as “Speaker 1” appears at its start. That label is what makes it useful for writing a story:

1. Click the label and type the real name. It changes in all of that person’s paragraphs at once.
2. If the engine split one person in two, give both the same name and they merge.
3. To find what someone said, type their name in the browser’s find on page (Ctrl+F) and move through the paragraphs.

At a press conference, that turns “what did the minister say to that question?” into a quick search. Short murmurs and overlapping speech can fool the separation, so check by ear before attributing words. On sources with a player, clicking a paragraph’s time takes the stream or file to that moment.

The whole team in one text

In a newsroom, the biggest gain is several people working on one text. When you share a transcript with your team, everyone edits at once and sees the others’ carets by name, and the faint text reaches every screen too. While a reporter follows from the field, an editor can work on the same lines in the office:

• The reporter starts the stream or microphone, follows what is said and fixes names and organisations.
• The editor writes the headline, orders the quotes and prepares the intro.
• A second reporter checks figures and past statements and adds notes.

If a change goes wrong, the Versions window takes the text back to an earlier state. Team sharing comes with team plans, as the pricing page shows.

A phone in the field, a strong computer in the office

In the field you may have only a phone. The small model on it works, but for more accurate text you can make a computer with a graphics card in the office your transcriber:

1. In the office, open the computer as a Transcriber device and leave the page open.
2. In the field, share the transcript with the team, choose that computer under Where to transcribe? in Settings and start the microphone.
3. The phone sends the audio straight to the computer over an encrypted connection, without touching the server. The office computer writes the text while the editor works on it.

If the two cannot connect (some networks do not allow peer-to-peer connections), the audio is held on the phone and the connection is tried again. So always test this setup before an important statement.

Before you publish

Speed does not replace checking. Automatic transcription can slip on negatives, numbers and names, so listen to every sentence you put in quotes. A detailed checklist is in our interview transcription guide.

Limits of live editing

• It does not turn the text into a story for you, summarise or translate.
• Faint text cannot be edited, so wait for it to become final.
• Tab sharing for live streams needs a Chromium-based browser on a computer.
• The free plan has 30 minutes in total, which may not cover one long statement.