Meetings

How to label speakers in a meeting transcript (and fix the wrong ones)

A transcript is only useful if you know who said what. How to name every speaker, fix mix-ups fast, and get cleaner labels on your next call.

Nov 17, 20265 min readBy Bruce W.

To label speakers in a meeting transcript, record each side separately if you can (your microphone as you, your computer’s sound as everyone else), then name each voice once and let the software apply that name to every line in the same voice. When a label is wrong, fix one line and check the lines around it. Spend your checking time on decisions, promises and questions, because those are the lines people quote later.

Why do transcripts say Speaker 1 and Speaker 2?

Speech-to-text software does two separate jobs. One writes the words. The other, called diarization, splits the recording by voice and decides where one person stops and the next starts. Diarization can tell that two stretches of audio sound like the same person, but it has no idea what that person is called. So it falls back on placeholders: Speaker 1, Speaker 2, Them.

The result is a transcript that has the right words in the right order but still can’t tell you who agreed to send the contract. That’s why speaker labels matter more than most people expect.

What makes speaker labels go wrong?

Most label mistakes come from four things:

  • One shared microphone. In a meeting room, every voice comes through the same mic, so the software has to tell people apart by sound alone.
  • Similar voices. Two people with similar pitch and accent get merged into one speaker.
  • Talking over each other. When two people speak at once, the overlap usually goes to whoever was louder.
  • Short replies. “Yes”, “sure” and “sounds good” are too short to match to a voice reliably.

Knowing the cause tells you where to look. If a transcript merges two people, check the stretches where they went back and forth quickly.

How to get better labels before the call

You can prevent most mix-ups before anyone says a word:

  1. Record your side separately. On a video call, an app that records your microphone and your computer’s sound as two tracks always knows which lines are yours. In a one-on-one call, that alone puts the right name on nearly every line.
  2. Wear a headset. It keeps other people’s voices from leaking into your microphone and being labeled as you.
  3. Say names at the start. “Hi Priya, hi Sam” gives both a person and an AI tidy-up something to work with later.
  4. Use names when you hand over. “Sam, what do you think?” tells everyone reading the transcript who speaks next.
  5. In person, put the laptop in the middle. We cover room setup in How to record and transcribe in-person meetings.

How to name and fix speakers after the call

Whatever tool you use, the same order works best:

  1. Name the biggest speakers first. Find a long line from each unnamed voice and give it a name. Long lines are the easiest to recognize.
  2. Apply the name to the whole voice. Most tools let you rename a speaker once and update every line. Don’t fix lines one by one unless you have to.
  3. Scan for the high-stakes lines. Read every decision, number, deadline and promise, and check the name on each.
  4. Fix the stragglers. Short replies and crosstalk are the usual leftovers. Fix the ones that change the meaning and leave the rest.
  5. Then write the notes. Summaries and action items are only as accurate as the names behind them, so label first and summarize second.

Which lines need checking most?

You don’t need a perfect transcript. You need the right name on the lines people will act on or repeat. Check these first:

Line type Example Why it matters
A promise “I’ll send the revised quote by Friday.” It becomes an action item with an owner.
A decision “Let’s go with the second option.” It shows who signed off.
A number “We can do 20 percent off.” Quoting the wrong person can cost money.
An objection “Legal won’t accept those terms.” It tells you who to follow up with.
A question “Can you share the data terms?” Someone is waiting for the answer.

If your notes end up in meeting minutes, the promises and decisions are exactly what goes on the page.

What if the transcript has no speakers at all?

Transcripts copied from somewhere else are sometimes one long block of text. In that case, split it into speaker lines first. An AI tidy-up can usually do this from context, like questions and answers, greetings and names. Then name the speakers as above. If you have the audio, transcribing it again with an app that separates voices gives better results than guessing from text.

How Fetch helps

Fetch records your microphone as Me and your computer’s sound as Them, so your own lines are always labeled correctly, with no bot joining the call. To name someone, click the ▾ next to a speaker in the transcript and choose Who is this?, or right-click any line and pick from Who’s speaking?. Fetch then learns that person’s voice, fixes every other line in the same voice, and recognizes them on your next call. After a call, Fetch can also name lines from context clues, like “thanks, Jane”, when Claude or ChatGPT is connected, and if only one other person was on the call, their name goes on every “Them” line. Each person gets their own color, and voice fingerprints stay on your computer. Importing a transcript instead? Pick Which speaker is you? so your lines show as Me. See Record your first meeting and Import a recording or transcript.

The short version

  • Transcripts say Speaker 1 because software can tell voices apart but doesn’t know names.
  • Record your side separately, wear a headset, and say names early.
  • Name each voice once and apply it to every line, then check promises, decisions and numbers.
  • Label first, then write the summary and action items.
  • Fetch labels your lines as Me, learns other voices after you name them once, and keeps voice data on your PC.

Questions people ask

Why does my transcript put two people under the same name?

Usually the two voices sound alike, they talked over each other, or they shared one microphone. Fix one of their lines by hand, and tools that learn voices will split the rest apart.

Can I label speakers in a transcript I already have?

Yes. If the transcript already has names, check them and fix any that are wrong. If it's one block of text, an AI tidy-up can split it into speaker lines first, and then you name each speaker.

Do speaker labels need to be perfect for meeting notes?

No. The lines that matter are decisions, promises and questions. Check who said those, and the summary and action items will be right even if a few small remarks are off.

Still scratching your head?

Send us a note and a real person will get back to you.

Contact support