Who said that? Answering the speaker-label question we hear most

The question reaches me almost every week, in nearly the same words: how do I know who is speaking when nobody says their name? It usually comes from someone preparing captions for a panel, a staff meeting or an event talk, and it carries a worry underneath. They fear that a wrong label is worse than no label at all.

That worry is fair, and it shapes the answer. A speaker label is a small promise to the viewer. It says that the words after it belong to one particular person. If the promise is broken, a viewer who cannot hear the audio is misled without any way of noticing.

So the short answer is this: you rarely need to know a name, but you must always be consistent about who is who. Names help, yet a steady, honest label such as Host or Speaker 2 serves a viewer better than a guessed name.

My first habit with any panel recording is to listen to the opening few minutes with no keyboard in reach. I am not transcribing yet. I am learning the voices: pitch, pace, accent, the way one person clears their throat before answering.

Most recordings give more clues than people expect. The moderator often thanks guests by name early on, or reads a short introduction. A guest may refer to the person who spoke before them. Someone may be addressed directly with a question, and the answer that follows is theirs.

I keep a small note beside the recording listing each voice with a plain description and the minute where I first heard it. That note becomes the key for the whole file, and it saves me from re-deciding the same question forty times over.

• Note the first minute each voice appears
• Write a plain description, not a guess at a name
• Listen for introductions and direct questions

Once the voices are separated, you choose how to name them. If the recording or the supplier gives you names and titles, use them exactly as supplied, with the spelling you were given. Ask for a short list before you begin; it spares you a lot of second-guessing.

When names are not available, use role labels that a viewer can follow: Moderator, Interviewer, Guest, or Audience Member. If two guests share a role, number them and keep the numbers fixed for the whole video. Speaker 2 must never become Speaker 3 halfway through.

Avoid labels built on appearance or assumptions, such as the man in the grey jacket. They age badly and can be wrong. A neutral role is always safer and reads more calmly on screen.

• Use supplied names exactly as given
• Fall back on stable role labels
• Keep numbering fixed from first caption to last

Panels are rarely tidy. People interrupt, laugh over each other and finish one another's sentences. Learners ask what to do when two voices speak at once, and my answer is to protect readability first and completeness second.

If one voice clearly carries the point, caption that voice and let the other fall away. If both matter, split them into two short captions in the order a listener would perceive them, each with its own label. A dash at the start of a line is a familiar way to show a change of speaker within one caption.

Where the overlap is pure noise and nothing can be recovered, a short sound note such as crosstalk is more honest than a muddled guess. Say less, but say it accurately.

Some recordings defeat you. A remote guest with a thin connection, an audience question taken from the back of a hall, a voice you heard only once. In those moments the right move is to label the role you can defend, such as Audience Member, and move on.

Do not leave a gap in the labelling and hope viewers will work it out. A caption with no label after a run of labelled ones suggests the same person is still speaking, which may be false.

If the recording is important and the doubt remains, flag the line for the person who supplied it. A short query listing the time and the question costs them a minute and protects the finished captions from a quiet mistake.

• Label the role you can defend
• Never drop a label mid-conversation
• Flag unclear lines with their time

Here is the routine I teach, trimmed to its bones. Listen first and build your voice key. Agree names and spellings with whoever supplied the video. Pick labels and fix them. Caption, switching labels only when the voice truly changes. Then do a final pass reading only the labels, ignoring the words.

That last pass catches more errors than any other check I know. Because you are not reading for meaning, a label that jumps to the wrong person stands out at once, and an inconsistent number is easy to spot.

Practise it on a short sample before you attempt a long event recording. Ten minutes of careful labelling teaches more than an hour of rushed work, and the habit carries over to every file you prepare afterwards.