How to Get a Transcript With Speaker Labels
In WhisperAI, it's one switch. Turn on Speaker Labels when you upload the file and every turn in the transcript says who spoke. Every plan has it, Free included. Type in the speakers' names too, and WhisperAI matches them to the voices: Dr. Smith and Maria instead of Speaker 1 and Speaker 2.
Already have a WhisperAI transcript without labels? You don't need to upload the file again. Re-transcribe it with labels on.
Speaker labels on an uploaded file, in four steps
- 1
Upload the file
Drop your interview, meeting or podcast recording on the upload page. Audio and video both work, up to 5 GB and 10 hours a file.
- 2
Turn on Speaker Labels
In the upload settings, switch on Speaker Labels. Leave Number of Speakers on Auto, or pick 2 to 10 when you know how many people talk.
- 3
Add the names you know (optional)
Under Speaker Identification, turn on Identify Speakers, choose Names or Roles, and type in up to 8 of them, such as Dr. Smith and Maria, or Host and Guest.
- 4
Check the labels
Click a speaker's label and choose Rename speaker to change that name on every one of their turns. Use Change speaker to move a single turn to someone else.
Exports. Select Speakers when you download TXT, DOCX, PDF or JSON, and the names come with the text. For subtitles, turn on Include speaker names in the SRT or VTT export: each caption starts with the name, like [Maria], whenever the speaker changes.
Speaker identification: names instead of Speaker 1
Speaker Labels tells the voices apart. Speaker Identification tells you whose they are: it matches each Speaker 1 and Speaker 2 to a name or role you gave it, like Host and Guest, when it's confident. Any voice it can't match keeps its number until you rename it.
Speaker Labels on
- Speaker 1: When did the pain start?
- Speaker 2: About two weeks ago, after the long drive.
- Speaker 1: Is it worse in the morning or at night?
Speaker Labels on, Names: Dr. Smith, Maria
- Dr. Smith: When did the pain start?
- Maria: About two weeks ago, after the long drive.
- Dr. Smith: Is it worse in the morning or at night?
Forgot to turn on speaker labels? Re-transcribe the recording
It happens. In your library, open the recording's menu, choose Re-transcribe with other settings, turn on Speaker labels and add the names if you have them. Then pick one:
- Replace current transcription swaps the old transcript for the new one, and it costs you no extra minutes.
- Create new transcription keeps the original and makes a separate recording with the new settings. It counts as a new recording in your usage.
Does Whisper do speaker diarization?
No. Diarization means working out who spoke when. OpenAI's Whisper model does speech recognition, speech translation and language identification. It writes down what was said, not who said it.
To add speakers yourself, WhisperX runs pyannote's diarization model on top of Whisper's output. You need a Hugging Face access token and have to accept the model's user conditions first.
OpenAI's own option is a separate model, gpt-4o-transcribe-diarize. It's deprecated along with whisper-1, and OpenAI will remove both from its API on February 26, 2027. OpenAI's speech-to-text guide shows speaker labels only for that model.
If all you want is a transcript of a file with each speaker marked, you can skip all of that. WhisperAI does it in the browser, with nothing to install.
Speaker labels in AI transcription tools, compared
Seven upload tools and how each one turns speaker labels on, as of October 28, 2026.
| Tool | How speaker labels get turned on |
|---|---|
| WhisperAI | Speaker Labels switch in the upload settings. Number of Speakers: Auto or 2 to 10. Names or roles: up to 8, typed in before you upload. |
| TurboScribe | Recognize Speakers, under Speaker Recognition & More Settings, before you transcribe. 2 to 8 speakers or Detect Automatically. |
| Descript | Automatic by default. Turn on Always ask before detecting speakers to give the number first. |
| Otter.ai | Automatic. Speaker identification runs after an import finishes and tags new voices Speaker 1, Speaker 2. |
| Sonix | Automatic. Speakers come back as Speaker 1, Speaker 2. |
| Happy Scribe | Automatic speaker detection on AI transcripts. |
| Rev | Labels appear in the transcript editor. Rev's help page says it doesn't guarantee them. |
Most of these tools label every file automatically. In WhisperAI you flip the switch, which also lets you tell it how many people are talking and who they are. Want labels on every upload anyway? With Speaker Labels on, choose Save as, name the preset, then pick Set as default in its menu.
More Whisper guides
For the broader picture of the Whisper model, see Whisper transcription explained. To skip the install and transcribe in a browser tab, see OpenAI Whisper online. To run it yourself, see how to download and install Whisper. If you're captioning interview video, Whisper SRT export covers subtitle files, and live transcription in your browser covers meetings as they happen. Comparing tools? See WhisperAI vs Otter.ai and the wider alternatives hub.
Frequently asked questions
Can I get speaker labels during a live meeting?
Yes. When you record live in WhisperAI, turn on the Label speakers live option in the session settings and the live transcript shows who is talking as people speak. That works in 18 languages. When you stop, WhisperAI analyzes the voices again, so the speakers in the finished transcript can differ from the live ones.
How many speakers can I tell WhisperAI to expect?
Set Number of Speakers to anything from 2 to 10, or leave it on Auto and let WhisperAI decide.
Is speaker labeling free?
Yes. Speaker Labels and Speaker Identification are on every WhisperAI plan, Free included. What changes between plans is how many minutes you get: Free has 5 minutes a month to try it, and Business Pro is unlimited for $24.99 a month.
Get a transcript that already knows who's talking.
Upload an interview, podcast or meeting, switch on Speaker Labels, and open a labeled, editable transcript.
See plans and pricing.