# Speaker labels transcript

URL: https://neattrim.com/tools/speaker-labels-transcript

Transcribe a recording and mark who is talking, so interviews and meetings read like a script.

## Transcripts that show who said what

A meeting transcript with no speaker names is nearly useless. When four people are discussing a budget, the question of who proposed the cut matters more than the exact wording. Interviews have the same problem: a journalist needs to attribute a quote, and a researcher needs to separate the interviewer from the participant.

This preset runs speaker detection and the transcript together. It listens for changes in the voice and starts a new labelled turn each time the speaker changes. Labels are generic, such as Speaker 1 and Speaker 2, in the order that voices first appear. Replacing them with names afterwards takes one find and replace in your document.

The quality of the labels depends on the recording. Voices that are clearly different and recorded at a similar volume are separated reliably. Overlapping speech is the hardest case. When two people talk at once, the words may be assigned to one of them, and a short interjection can be missed. A recording with one microphone per person, where available, gives better results.

The cleanup step and its edits are not applied. The result is a transcript of the recording exactly as it was.

## Steps

1. **Add the recording.** Drop in an interview, meeting, call or panel as audio or video.
2. **Keep speaker detection on.** This page turns speaker labels on and exports only the transcript. No video or audio is returned.
3. **Set the spoken language.** Choose the language if it is not English, so that recognition uses the right model.
4. **Choose TXT, SRT, VTT or DOCX.** Take text or Word for reading and editing, or subtitles if you want the labels on screen.

## Questions

**How many speakers can it separate?**

It can separate several voices. Accuracy drops when voices are similar or when more people join a loud discussion.

**Can I get real names instead of Speaker 1?**

The labels are generic. Replace them with names in a text editor using find and replace.

**Will overlapping speech be labelled correctly?**

Overlaps are the weak point. The words are usually kept but may be attributed to one speaker.

**Which format keeps the labels with timestamps?**

SRT and VTT carry timing for each turn, and DOCX and TXT are best for reading.
