Plaud Speaker Identification: Set Up Labels and Check Who Said What
Set up Plaud speaker labels, name voices from audio, correct attribution errors and check how personal voice profiles work across recordings.
Plaud speaker identification separates voices in a transcript. With your own voice profile added, Plaud can label your speech with your name. Other people appear as Speaker 1, Speaker 2 and so on. You still need to listen and assign their names. A readable transcript can attribute a perfectly sensible sentence to the wrong person.
We checked Plaud's official App and Web guides on October 25, 2026, including its newly separated instructions for v4.0.x and v3.30.x or earlier. The steps below come from those guides. We have not tested a signed-in Plaud account or recorded a voice profile for this article. Our PLAUD Note product page covers the hardware; this guide follows the speaker labels after transcription.
Speaker Labels Are Not Verified Identities
Speaker diarization means separating speech into speaker groups. Speaker 2 collects passages that the software assigns to one voice. You can give that group a name, then check the audio to confirm the person actually spoke those words.
Plaud's current automatic labeling instructions say automatic recognition supports your own voice when you have added your voice profile. Names you manually give other people apply to that recording only. Do not expect a colleague named in Monday's interview to receive that name automatically on Tuesday.
A linked calendar event offers a convenient list of names, but you choose which attendee matches each voice. Plaud explicitly says it does not make that match automatically. Someone can appear on the attendance list without speaking at all.
Start with an audible introduction, a voice you confidently recognize, or confirmation from someone who attended. A job title, subject expertise or an apparent assignment in the transcript is too weak a basis for attaching a name. Keep an uncertain speaker's original anonymous label and note the uncertainty in your review notes.
Set Up Automatic Speaker Labeling
Check your App version before following a menu path. In Mobile App v4.0.x or later, Plaud documents the following setup.
- Open the top-left menu on the home screen, then open Settings.
- Under Workspace, open Preferences, then Transcription & summary, then Auto speaker labeling.
- Enable Label my voice automatically before generating a transcript.
- If you want Plaud to name your own voice, select Add your voice profile, name it, and record at least 30 seconds of speech after granting microphone access.
Despite the switch's reference to your voice, Plaud says it controls speaker labeling for new transcripts as a whole. With it off, new transcripts have no speaker labels, including yours. Other participants do not need voice profiles for you to name them within a recording.
For v3.30.x or earlier, the older App guide puts the setting under Settings, Preferences, then Auto speaker labeling.
On Plaud Web v4.0.x or later, click your workspace name, open Settings and Preferences, then find Auto speaker labeling in the Generation section. The Web setup guide describes the same labeling switch and optional personal voice profile.
Turning the switch on will not add labels to a transcript you already generated. For an old file, save your edits first, then re-transcribe with speaker labeling enabled. In the current Mobile App, the recording's top-right menu offers Regenerate, followed by Transcript and Summary. Check the speaker labeling setting before selecting Generate.
Plaud's regeneration guide says this replaces the current transcript, summary and notes. It also removes your edits. Re-transcribing is free of transcription minutes and Plaud Credits according to that guide, but copying your work beforehand still matters. Re-running the same unclear audio rarely improves its quality.
Review and Correct the Transcript
In the current Mobile App, open the recording and tap the Speakers chip in the header. Plaud's meeting info and speaker labels guide describes a naming window with speaking times and playable transcript snippets. Listen to those snippets, type the names you can support, then tap Confirm. The top-right menu's Name speakers option opens the same window.
On Web v4, click a speaker chip in the file's Speakers row or a label in the transcript, enter the name, then select Save. Its naming guide says names update throughout the file, including the summary. Check several snippets before applying one person's name to a whole speaker group.
Use this table to decide where a rename helps and where listening must continue.
| Symptom | First check | Supported correction | What still needs listening |
|---|---|---|---|
| The wrong name appears | Compare several clips with an introduction or a known voice | Replace the name in the naming window | Every important statement attributed to that group |
| Two people share one label | Listen around each change of voice | In the older App, apply a corrected name to an individual paragraph. The current guides do not document a universal split-speaker tool, so mark unresolved passages in review notes | All mixed passages, especially interruptions |
| One person has two labels | Confirm both groups sound like the same known person | Correct the verified groups' names. The current App guide also mentions merging speakers, without documenting its exact controls | Other passages in both groups before treating them as one person |
| An old transcript has no labels | Check whether labeling was enabled when it was generated | Copy existing work, then re-transcribe with labeling on | The newly generated groups and names |
The older App naming instructions distinguish applying a name to one paragraph from applying it to all paragraphs with that label. Use the scope appropriate to the evidence. They also say an existing summary needs to be generated again after renaming. Current v4 App and Web instructions say renaming updates the summary automatically. Re-read the summary after either process.
A fictional example of checking who said what
The following short exchange is invented for illustration. It is not a Plaud test recording.
- Speaker 1 says, “I'm Maya. I'll send the revised quote on Friday.”
- Speaker 2 says, “Leo here. Please leave the delivery date open.”
- Speaker 1 says, “Agreed. The quote will list 24 units.”
The introductions provide a starting match. Listen to the third utterance to confirm that it has Maya's voice before attaching her name to it. If Speaker 1 actually contains two voices, renaming the entire group Maya would spread the error. Keep the uncertain passage marked for review instead.
Before sharing minutes, replay the commitments and numbers. In this example, check who promised Friday's quote and who said 24 units. Give sensitive attributions the same attention. An AI recorder can help you find who said what; use the audio to support the name before quoting it.
Decide Whether Labels Should Carry Across Recordings
The current setting Sync voice profile across devices concerns your own voice profile on your signed-in devices. Plaud says turning sync off removes that profile from the cloud. You can also delete your profile in the labeling settings. Choose whether to use these options after reviewing the settings available to your account.
Names you enter for other participants apply to that recording only. This naming rule says nothing about where the audio file is stored. Check your file storage settings separately.
Check Plaud speaker labels again for every recording. On Web, an assigned name can be replaced, but Plaud says it cannot simply be cleared back to the original default label. Avoid assigning a speculative name in the first place. Re-transcribing resets manually assigned names, while your own saved voice profile can still label you automatically.
If you are still choosing a recorder for meetings or personal recall, our AI note takers comparison covers the broader device choices. Leave any unresolved identity visible in the review notes you share with the transcript.