Premiere 24.0.3 has trouble identifying the correct speakers

el hoozn,

Odd isn't it? I looked more closely, and you can see from info in the tutorial that she created this 5/7/23 in a Beta version of PR. So I don't think there's any reason to suspect a change as of 24.0.3. And, as I said, I am seeing the same thing in the current Beta version of PR.

The clip used in the tutorial is different from the one provided in the project. The clip she used is longer. There, transcript segment 1 begins with the interviewer explaining the setup and ends with what we first hear. The Speakers are identified correctly. My guess is that, with a longer sample of the first speaker, PR is able to correctly identify transcript segment 2 as from the same speaker and to identify Speaker 2 correctly.

Thanks for reporting this. Other users will find this. And workarounds often pay off at unexpected times.

Stan

Hi @Kerstin Ebert 

Tbh I dont intend to work with those interview files, so I dont actually need a workaround right now, but thanks, I guess.

I just noticed that using the video snippet provided by Adobe somehow leads to a different outcome than it should (according to the result seen in the tutorial). Assuming that said tutorial has not just been created for this particular update (24.0.3), and assuming that the material used for the tutorial is the exact same as the one provided to the users, the difference between expected behaviour  and the actual result would hint to an error that has been introduced somewhere between 24.0 and 24.0.3.

Or is getting different results for the same transcription actually expected behaviour?

greets

@Kerstin Ebert,

Thanks for commenting. I forget how easy it is to reassign individual transcript segments when they are misidentified.

This result is worrisome. I get the same results transcribing in 24.0.3 and Beta 24.2.0.15. I am assuming the "Interviewer" is also operating the camera. We see him briefly when he clicks the slate and says the first line - in the same location as Billy. When he says his second line, he is now located at the camera, and sounds distant because the sound is being recorded via lavalier on Billy.

Transcript segment 1 is identified as Speaker 1. Transcript segment 2 is identified as Speaker 2, but is, I believe actually the Interviewer. It may be speaker 2 if a camera operator is handling the slate and speaking that first line.

Transcript segment 3 is identified as Speaker 1, but is in fact, Billy's first line.

@el hoozn, if you actually had this situation, the easiest fix would be to correct this before renaming - just reassign the first transcript segment to Speaker 2. Then name them with Billy as Speaker 1.

Stan

Hi @el hoozn,

Thanks for your feedback! It looks like the transcription enginge can't clearly identify the speakers in the very first text segment, most likely due to background noise. I think the easiest solution would be to switch the speaker of the 2nd text segment ("All right, let's start with you telling...") from Billy to Interviewer – just click on the three dots next to the speaker name, this opens a speaker drop-down. Now speaker 1 can be renamed to Billy and speaker 2 can be renamed to Interviewer. This way the speakers should be assigned correctly. Does that work for you?

Best,

Kerstin