Works out where the speaker changes in an interview or panel transcript and labels each turn. Inferred from the words, because we never see the audio.
Or pull it from a video
Paste an interview transcript. Turns are labelled from the words, not the audio.
This is an example of what Speaker Label Generator returns. Run it above on your own input to get your own result.
Video: Podcast: opening a bakery with Priya Nair 0:00 Welcome to the show. Today I'm joined by Priya Nair, who opened her bakery last spring. 0:09 Thanks for having me, Tom. It's been a busy year. 0:14 What made you leave your office job? 0:19 I got tired of staring at spreadsheets. I'd been baking bread at weekends for years. 0:28 And how did the first month go? 0:33 Slow. We sold eleven loaves on the first day and I cried in the stockroom. 0:41 When did that change? 0:44 A woman bought six rolls and came back the next day with a friend. After that, it grew by word of mouth.
Video: Podcast: opening a bakery with Priya Nair 0:00 Tom: Welcome to the show. Today I'm joined by Priya Nair, who opened her bakery last spring. 0:09 Priya: Thanks for having me, Tom. It's been a busy year. 0:14 Tom: What made you leave your office job? 0:19 Priya: I got tired of staring at spreadsheets. I'd been baking bread at weekends for years. 0:28 Tom: And how did the first month go? 0:33 Priya: Slow. We sold eleven loaves on the first day and I cried in the stockroom. 0:41 Tom: When did that change? 0:44 Priya: A woman bought six rolls and came back the next day with a friend. After that, it grew by word of mouth. Note: speaker labels are inferred from context and phrasing, not from audio, so a hand-off between speakers may be mislabelled.
Paste a multi-speaker transcript and we label who is talking, Speaker A, Speaker B, or a name if one is obvious from what's said, in front of each line. Any timestamps already in the text stay exactly where they are.
We never hear the audio. Labels are worked out from the words alone, so a fast back-and-forth or a hand-off mid-sentence can get mislabelled. The result says this plainly at the end rather than pretending the split is certain.
No. We never receive the audio. Labels are inferred from the words and the flow of the conversation.
Good on a clear back-and-forth, weaker on fast interruptions or a hand-off with no verbal cue. Check it against the recording if the speaker matters.
Yes. Speaker Labels is a Starter tool.