Audio Track Selection for Source Media Transcription

Report · May 30, 2023

Edit 10/30/23: Update on Multichannel Audio Transcription Support. PR Release version 24.0.0 added transcription of multiple audio channels for source media, but only for mono. Stereo has been added in the Beta version, and is in active development.

Auto-transcription still looks only at audio track one.

If PR recognizes the file as having multiple channels, it adds "Channel Selection" to the options in the "Create Transcription for Source Media" dialogue. If it does not see multiple channels, you do not see that option.

Some MXF files with multiple mono channels are not yet working correctly.

See these posts from staff:

https://community.adobe.com/t5/premiere-pro-beta-bugs/text-based-editing-doesn-t-work-with-multiple-...

https://community.adobe.com/t5/premiere-pro-ideas/audio-track-selection-for-source-media-transcripti...

Original Post: Text-Based Editing/Source Media Transcription as rolled out in Releaase 23.4.0 does not allow for the selection of which audio track to transcribe, but transcribes only track 1. From forum posts, users often appear to have multiple audio tracks per file, including microphone, camera(s), sound board input, etc. The workarounds quickly become onerous.

I recognize the difficulty, which is encountered when using nesting workaround. Instead of one transcription per clip, you must have multiple transcriptions per clip. In my experiments, the audio track picked currently for transcription is not changed by mapping audio channels etc. So any method using multiple audio tracks/transcriptions would need to name them in a way that the user could understand and track.

Workarounds:

Modify the source file so the desired clip is audio track 1.
Place the clip in a sequence, create a static transcription (which allows audio track selection), and then nest that sequence. That nested sequence carries the transcription and behaves like a source transcription.
Select the multitrack file in the Project Panel and Clip -> Audio Options -> Extract Audio. This creates as many files as there are audio tracks, named [Filename] + Extracted.wav, [Filename] + Extracted_1.wav, etc. Note that the numbers are in revers order to the audio tracks in a sequence. For example, if there are 2 audio tracks, the file named ...Extracted_1.wav will be audio track one and the file named without the _1 will be track two.

Note that anytime transcribed clips are on different tracks in a timeline, the higher-ranking track (e.g. audio track 1 is higher than audio track 2) will be active.

If you are upvoting this idea, be sure to click the numbering box in the upper left of the thread, not just individual posts.

Stan

Report · Jul 26, 2023

@Stan JonesWhat woudl be the best way to modify the source file so the desired clip is audio track 1? Fortuantely I only need to dealw ith this issue for 4 files, here's a sample:

Report · Jul 28, 2023

@Mauronic,

That is a good question. Thanks for the screenshot. The problem is I have not played with enough actual workflows and file types, and I don't know whether some of the workarounds have issues later in the editing.

I suspect #1 works best, but only if you have a lossless option to reformat the source file before brining it into PR.

#2 is what I would pick watching carefully for any issues later in the workflow.

#3 is still hard for me to get my head around, and may ultimately involve the sort of nesting required by #2.

Stan

Report · Jul 30, 2023

Just came across a similar issue. Dealing with source files with multiple language commentaries, A1&2 is clean fx, A3&4 English, A5&6 French etc.

Subclipping and/or adjusting audio mapping has no effect - only audio on channel 1 of the source file is transcribed.

Workaround is to put into sequence and do static transcript - which works pretty well TBH but it would be great to be able to transcribe multiple languages.

Report · Jul 31, 2023

Perhaps there is another tool out there that would allow you to manipulate the audio outside of premeire so you can do #1 lossless.

In the example above, I started with Tentacle Sync Studio to sync my Tentacle Track audio recordings to my movie clips using their timecode solution. I used their export to MOV option witch I beleive left the video track untouchaed and just packaged the audio with it.

As an aside, I just used their software's XML export option, instead of the export media option (MOV). When I opened the XML file, everything came into premiere exacly how I needed it. So I have a workaround for my unqiue use case.

Report · Jul 31, 2023

I'd love to get @Jarle Leirpoll in this discussion. Joost Van Der Hoeven also but I don't know his moniker to ping him.

Report · Aug 01, 2023

@Mauronic,

Thanks for the additional thoughts about workarounds.

XML has some powerful workaround options for many functions.

Stan

Report · Aug 06, 2023

@R Neil Haugen I think @Stan Jones has laid out the problem and the workarounds perfectly. We absolutely NEED a way to tell Premiere Pro what audio track to use for transcription.

The current solution works fine for people who record using their phones or system cameras (usually "stereo" sound with the same audio on both channels, so in practice, just double-mono), but for those who use professional cameras with 4 audio channels or more, or separate audio recorders with multiple channels, it's not working great, and sometimes not at all.

Report · Aug 06, 2023

Thanks for popping in, Jarle!

We've needed more and better user options for setting up audio for some time. As you have commented on several occasions.

Between proxy audio, channels for specific broadcast needs, and others, it's more potentially complicated these days.

Report · Aug 18, 2023

Hi Stan, and thanks for bringing the attention to this limitation and helping other users.
I have tried #1 but it has not worked. It seems as if there is a cache for the attempt that failed which prevents it from seeing the "new" audio on A1. It's a guess.

I have tried #2 and #3 several times with no success either.
I will continue using static transcriptions which are already a major help until the bug is fixed.

Rubén

Report · Aug 30, 2023

I edit a weekly podcast. The interview is typically a nested sequence with the interviewer on A1 and the guest on A2. When I run the transcript for the finished sequence, it only displays text for A1. This is incredibly frustrating, and replacing it with a static transcript defeats the entire purpose of the linked transcript system. Hopefully this gets fixed sooner than later.

Report · Aug 31, 2023

@avid314786382y9u,

Your workflow is a great example of the challenges in text-based editing. Even if source media transcription allowed specifying which tracks to transcribe, you'd still have the issue described in my note above: "Note that anytime transcribed clips are on different tracks in a timeline, the higher-ranking track (e.g. audio track 1 is higher than audio track 2) will be active." From your description, I assume you have separate audio for the speakers, so you can transcribe them as independent sources. BUT, when you add them into a sequence, one will always be in the higher audio track, and the other's transcript will not be visible. To test this, mute the higher audio track, and you will see the lower track's transcript.

I think the best option for you is to do a static transcription of your sequence with both audio tracks, using the identify speaker and "mix" settings. This creates a transcript of both speakers, hopefully with the speaker designations being correct.

Now nest that into a new sequence. The nested sequence (and its "static" transcription) are treated as a source transcript in that new sequence. That transcript will in fact update as you edit, and you see both speakers for editing.

You'll have to manually deal with any times the speakers talk over each other.

I'm wondering what the right feature request is for such situations. A toggle to show all transcripts (e.g. those hidden by higher tracks)? But how do you display those when they overlap others; columns of tracks? And how would you edit them? The current method (one transcript and using mute as a workaround to show a hidden track) is limited, but what is the option?

Stan

Report · Sep 20, 2023

"We absolutely NEED a way to tell Premiere Pro what audio track to use for transcription." Of course! How is it that Adobe hasn't figured out this most BASIC functions. And why does this feature exist in static transcript and not when transcribing the source?????

Report · Sep 20, 2023

Yes, please fix this. This is so bad and causes so much wasted time and effort. Please allow selection of audio track asap.

Report · Sep 23, 2023

They are starting to work on this in the public Beta. The "What's New" note in the PR Beta says "Source Multi Channel" and "Enables multi channel options for source clip transcription in the text panel." I thought I saw a staff post, but I cannot find it. In any event, I see no new option enabled in the Beta, so probably early days for its development.

@TeresaDemel @Annika Koenig,

Perhaps I just dreamt that one of you posted about this. lol. Is there any update posted?

Stan

Report · Sep 26, 2023

Hi @Stan Jones,

I'm glad to say that you didn't dream this, we just released multi-channel transcription in Beta!

If you want to transcribe only 1 specific channel of your multi-channel audio, you can double click on the source clip and then click on "Transcribe" in the Text panel (or: right-click on source clip in Project panel and select "Transcribe". This will open the transcription settings where you can select the channel you want to transcribe in a drop-down menu.

It works differently when you have auto transcrip tion on – then we transcribe channel 1 as default.

Let me know if you still don't see it.

Thanks,

Kerstin

Report · Sep 26, 2023

@Kerstin Ebert,

Thanks! I don't see it in Build 32. Will keep looking in the next build.

Stan

Report · Sep 27, 2023

I'm glad to see that track selection for source transcription is in the Beta. This is very much needed.

Regarding the workaround, if you do a static transcript for a sequence, you can load the sequence into the source monitor and the text-to-edit window will allow you to follow along and make selections. This method doesn't solve features in the resulting edit, but it does allow you to use the text tools to find sound bites and add them to an edit, which is what is needed most often at my workplace.

In general, Premiere's handling of MP4 audio (single stereo clip) VS multichannel audio is a real problem, especially when you layer the Frame IO "workflows" on top of it all. I don't want to see Premiere pre-render all incoming audio to a multichannel standard, but if that's what is takes, maybe that's what it takes.

Report · Sep 27, 2023

@Kerstin Ebert,

This second part of this announcement from Teresa was my "dream":

https://community.adobe.com/t5/premiere-pro-beta-discussions/now-in-beta-filler-words-and-other-text...

Neither of these functions (filler word selection in transcription; multichannel audio selection in source media transcription) are showing up in Build 32. Any idea when they will be up?

Stan

Report · Sep 27, 2023

Hi @Stan Jones,

all of these features should show up in the latest Beta build. If you're not seeing them, I have a few questions for you:

1. Which OS and hardware are you using?

2. Did you double check if fillers are turned on? You can do this in the "Transcript view options" (in ther ... menu, or in the little icon in the bottom right of the Text panel). It looks like this:

3. How are you trancribing your multichannel audio? Is it a mono file?

Thanks,

Kerstin

Report · Sep 27, 2023

I've installed the current beta 24.1 and do not see options for audio track selection when transcribing.

Report · Sep 27, 2023

I wonder about my files. Can you provide a link to a multichannel sample that works?

Fillers is on.

i7 11700K 3.6GHz 8 Core 16MB
Installed RAM 64.0 GB
Windows 10 Pro 22H2 OS build 19045.3448
NVIDIA GeForce RTX 3070, 8GB VRAM, Studio Driver 536.23

Stan

Report · Sep 28, 2023

@Kerstin Ebert,

Updated to Build 35, which has a bug that prevents access to the Create Transcription dialogue for source media.

https://community.adobe.com/t5/premiere-pro-beta-bugs/transcribe-source-reload-text-panel-error-24-1...

Any way around that? I'll just wait for the new build to test. If you have a test file you can post that will help.

Stan

Report · Sep 28, 2023

I also have the "no detected dialoge" problem with audio transcibing, i have 4 audio channels and channel one has no audio, i have the latest beta build and i am on a macstudio ultra running OS version 13.5.2.

Report · Sep 29, 2023

Hello @Xzault

it sounds like channel 1 was transcribed because it came back empty. Do you have auto-transcibing turned on? If yes, that you won't see the "Transcribe source media" dialogue which let's you chose which channel you want to transcribe. Instead it just transcribed channel 1 as default.

If you want to manually select which audio channel should be transcribed, you have to turn auto-transcription off, open the source in the Text panel, click on "Transcribe" and select the channel from the drop-down in the transcription dialogue (or: right-click on clip in Project panel and click "Transcribe").

Please let me know if that still doesn't work.

Best,

Kerstin

Report · Sep 29, 2023

I have auto transcribe turned off, it does not give me that option to choose my audio channel, it only wants me to chose language and speaker labeling options. MAybe im missing something ?

Adobe Community

Audio Track Selection for Source Media Transcription