Transcripts are much easier to read when the speaker changes are reflected in the transcript with paragraph breaks, and those paragraphs are correctly labeled with the speakers' names, when known. Reduct's algorithm attempts to detect when the speaker changes, and when different speakers' voices are present, and automatically labels paragraphs with "Speaker 1", "Speaker 2", etc.

By clicking on the speaker name, (whether it's "Speaker 1" or if it's been previously renamed), you can rename it. The algorithm identifies the same voice in multiple portions and marks it with the same label. While changing the speaker label, it will give you the option of renaming all paragraphs that have been identified as that same voice.

Check the option to “Change X instances” to rename all the paragraphs with your entered speaker name.

The speaker identification is particularly sensitive to audio quality, and even with the best recordings can sometimes confuse (or just not differentiate between) speakers who sound similar (which most often happens with two speakers of the same gender).

While editing the speaker name, a checkbox usually appears stating ‘Change X instances’.

If you uncheck this box, you can override the system's interpretation of where a particular speaker is talking. Once you name multiple paragraphs with the same speaker name you can act on them in a single go.

In the example below, I want the third paragraph to be named "Kate Williams" as it's not the same speaker as the others marked "Speaker 2", so I uncheck the box. In the second action, I want to rename all remaining paragraphs labeled "Speaker 2" to be labeled as "Yvon."

If you rename any two paragraphs with the same speaker name, they will be considered to be the same speaker. This also applies when you rename "all instances" of a speaker.

So, for example, you might have a transcript with three speakers identified (Speaker 1, Speaker 2, and Speaker 3), but there are in actual fact only two voices. If one of those voices is sometimes marked Speaker 2 and sometimes Speaker 3, you can rename them to be the same and then the system will treat them as a single speaker.

If the challenge is that the breaks aren't consistently right, then the first step is to manually break apart paragraphs where you hear the speaker change (and then optionally, recombine them).

In the example below, there's a sentence at the end of the first paragraph that is actually the second speaker, so I select it, "extract paragraph", rename it to match the ones I want it to belong to (making sure the "change all" box is unticked), and then merge it with the following paragraph.