How do I get separate transcriptions for each speaker when they are on two separate audio tracks.

[bobw111]

I have Premiere Pro video with two speakers and each speaker is on a separate audio track. What is the best way to get each audio track transcribed separately?

I tried asking Adobe's AI chat system and it is absolutely useless.

[Stan Jones]

@bobw111,

Are these in their own separate files? If a single file, right-click the file in Project Panel and pick "Transcribe." Pick Re-transcribe if you already transcribed it. Do you see a "Channel Selection" option? Do you see the channels you expect?

If any of this does not appear as you expect, post back.

If they are in one file, you can have only one transcript at a time. So I use Clip -> Audio Options -> Breakout to Mono. That does not create files on disk, but they will keep their transcription. Note that, if you already transcribed the original clip, each such breakout will start with that transcription. Just pick re-transcribe.

Stan

[bobw111]

It is an OBS recorded MP4 video file with two separate audio tracks (A1 and A2 in Premiere).  Each audio track has left and right channels.  I found the setting in premiere to separate individual speakers transcription text, but it is in the "auto transcribe" section and I don't want a transcription on every video I import.  I'll do a bit more testing with it later today.  Maybe I'll just have to turn on auto transcribe for this video, then shut it off again.  One other irritating thing I found is there doesn't seem to be a way to delete a transcription and create a new one without clearing cache.  If you delete the clip from the time line and then re-import it PrPro automatically grabs the old transcription.  I thought I found the way to turn that off, but it seems to only apply if that video is brought into a totally different project.  I'll post the results of my testing.

[bobw111]

Hmmm... Clip-> Audio Options -> Breakout to Mono is grayed out.  I've also noticed that "Re-transcribe" is always grayed out.  Guess it's time to watch a bunch of videos and try to figure out why my mileage is different...

[Stan Jones]

> Clip-> Audio Options -> Breakout to Mono is grayed out. I've also noticed that "Re-transcribe" is always grayed out. Guess it's time to watch a bunch of videos and try to figure out why my mileage is different...

Be sure you are selecting the media FILE in the Project Panel and not a sequence.

> PrPro isn't doing a very good job of identifying speakers. It seems to confuse the voices of the separate speakers even though they are on separate audio tracks.

I sent you a private message. I'd be happy to test a clip. There can be a lot of variation.

I forgot to ask: what version of PR are you using? Mac or PC?

> create two separate projects for the video clip

There are other ways around this; no need for 2 projects.

Stan

[bobw111]

After some more testing I've noticed that PrPro isn't doing a very good job of identifying speakers.  It seems to confuse the voices of the separate speakers even though they are on separate audio tracks.  I think the only way I am going to get a reliable transcript is to create two separate projects for the video clip, using audio track A1 on one project and audio track A2 on the other project.  Then transcribe each project and export the resulting text out of each project as a csv file.  Then use Excel to import the two CSV files in, then sort the excel file by time stamp (possibly having to add a column in each csv file to identify speaker by.   Seems like a lot of work when I've got each speaker on a completely separate audio track.   Separate tracks for each speaker should make the transcription process easier, not harder.

[Stan Jones]

@bobw111,

First, I'm going to post what I had drafted earlier, but only finished just now. Then I'll address your new posts.

Turn OFF autotranscription.

> I found the setting in premiere to separate individual speakers transcription text,

You have the same selection in non-automatic transcriptions. What this does is to try to detect each speaker, and to label them (Speaker 1, etc). This can work for you, and has some advantages over separate transcriptions by audio channel. try that first.

I find it easiest to control what is happening by transcribing from the Project Panel.

Right-click your clip in the Project Panel, and pick MIX from the audio channel options and "separate speakers." Mix looks at all the audio channels in the source clip. As long as the 2 speakers do not talk over each other, it should separate the speakers well. Double-click the clip in the Project Panel (so it opens in the Source Monitor), and switch to the Text panel/Transcript tab. Edit the speaker names there. (You cannot edit speaker names of a source media transcript once in timeline/sequence view.) And go ahead and edit the content so each speaker is associated with their content. You can split segments as needed. There will always be some editing needed for speakers, but getting separate transcriptions by speaker combined into one will have other problems.

This will create a single transcription with both speakers.

Stan

[bobw111]

Sorry I'm a day late following up.  I see one problem, despite trying to avoid it, the speakers do talk over each other once in a while.  It's late tonight, and I've got to get some recording done tomorrow for my Patreon & Youtube channels, but I will go through your recommendations and try them out as soon as I get a chance. Probably won't be until 10/3 or 10/4.  Fortunately this particular project is no rush.  BTW, I really appreciate you help and input.

[Stan Jones]

@bobw111,

No problem. Getting this workflow sorted will pay off in the long run.

With the overlap, you can edit the single transcript as needed to get what seems best for this.

The problem with multiple transcriptions is that you can only see the one for the clip higher in the audio track order.

Now that multiple caption tracks can be enabled on screen, this is not an issue for burned in captions, and you can find ways around the text-based editing challenge.

Stan