text based editing has serious flaws

[gabrielk41156439]

I've been using text based editing on a project which involved recording a workshop with two cameras and four audio recorders.  The conversations also involved small break out groups which were audio recorded. I imported each recorder file, transcribed them and created subclips from in and out points(clip/make subclip).  It's been very problematic and here's a list of the issues:

- The timecode of the subclip and the transcript is different and won't playback from the text, only from the video window and the text will not be follow with playback. 

- when I export the subclip transcript I generally, and by generally I mean 90% of my exports and completely randomly, I get an export of the full source file transcript and not the transcript for the subclip section although it appears on the text section of the subclip at such.  Additionally, eventhough the subclip transcript exports include the whole source clip transcript, the timecodes associated in the transcript coincide with the timecodes of the subclip.  So for example if a certain subclip begins at 07;07;07, then the transcript export gives me the full source transcript, but the timecode says 07;07;07 at the beginning.  This changes with each export.  It very very frustrating and wasting huge amounts of my time and resources. 

- when working on fixing a transcription from a sequence which includes multiple microphone channels from separate recorders , which in this case I used to capture a large group conversation, I don't know which source file I'm working with, meaning that I edit the transcript on the sequence but dont know if the changes are being copied to every source file on the timeline or which one is being actually rewritten aside from the sequence transcript. 

I have more bugs but these are by far the most annoying and the ones making me seriously consider not using premiere. 

[Kylee Pena]

Hey Gabriel, sorry to hear you're having some issues. Would you be willing to send me the .prproj file (no media) to look at? My email is kpena @ adobe.com. 

When you say you're working on fixing a transcription from a sequence with multiple mic channels, you don't know which source file youre' working with. You should see the filename in the upper right corner if you're on the source. If you're in the sequence, you'll see the sequence name. The changes you make to the transcript (such as correcting spelling) in either place will ripple through. I might be misunderstanding, so feel free to also send a brief screen recording of the issue. In any case, it's just not working right for you which is a bummer. 

[Stan Jones]

@gabrielk41156439,

Subclips generally do not behave as you expect. See this post for some discussion:

https://community.adobe.com/t5/premiere-pro-bugs/text-based-editing-sequence-doesn-t-match-clips-seq...

Some of the issues are:

• If you do not apply the "Restrict trims to subclip boundaries" the transcript of that subclip will be the transcript of the WHOLE clip.
• If you do not convert the subclip to a source clip (by editing the subclip once created), the timecode of the transcript will not work in a sequence the subclip is used in, and you will see "No dialogue."
• If the clip has been transcribed before, you must re-transcribe. You CANNOT use a previous transcription; the timecodes will be wrong.

Some of the oddities of timecodes are probably a result of the multiple options for settings timecode in subclips.

> when working on fixing a transcription from a sequence which includes multiple microphone channels from separate recorders , which in this case I used to capture a large group conversation, I don't know which source file I'm working with, meaning that I edit the transcript on the sequence but dont know if the changes are being copied to every source file on the timeline or which one is being actually rewritten aside from the sequence transcript.

Yes, difficult to keep straight. I'll have to come back later to comment on that part.

Stan

[gabrielk41156439]

Thanks for your reply, Kylee.  I've sent you my project file without media. With regards to the sequence/source file transcription issue. I've created a screen recording and sent via email.  thanks. 

[gabrielk41156439]

Thanks Stan.  So to understand what a workflow would look like with the characteristics of my project:

- Import media

- Create a sequence and sync (and clean) all audio files of the group discussion recording

- Make that sequence into a source file

- Create a new sequence

- place both video files and mixed recording source file in the timeline

- sync

- edit and create any subclips needed

- make them source clips

- transcribe

- fix transcription (one by one)

- export

If this is correct then it would seem as though text based editing isn't really a thing as I would need to retranscribe any edits I would like to turn into source clips for export, losing the changes I made in other versions/edits or source clips I used to create the subclips. If the auto transcriptions were perfect, which as they stand are not, then it would be easier. 

I hope there's an easier and more intuitive way that i'm not seeing at the moment. Thanks again. 

[Stan Jones]

@gabrielk41156439,

Thanks for that workflow outline.

Unfortunately, I have discovered that there is an error in my proposed workflow. When the subclip is converted to source, it does not just change the subclip times. I do not see way to do what I was trying to accomplish.

@Kerstin Ebert @Alexander_DVA

Can you guys help with a way to do what Gabriel wants?

Is there a way to create a subclip with a transcript that is limited to the subclip and shows up when added to a sequence?

I know my lack of knowledge about subclips is a problem here, but I've tried everything I can see.

Stan

[Kerstin Ebert]

Hi @gabrielk41156439,

thank you for sharing your detailed workflow! We are working in improving the behaviour with subclips that contain transcriptions. I don't have a perfect solution for you yet (we're on it!) but here is a workaround that might fix your problem for now:

- transcribe your source clips first, before creating subclips

- when creating subclips, select "unrestricted subclip" in the dialogue

- now when you add the subclip to your timeline, the timecode of the transcript should match

- if you want to open the subclip in the source monitor, do this by double-clicking on the subclip in the timeline (note: if you open the source clip via double-click in the Project panel, the timecode breaks - this is a known issue we're working on)

Let me know if that helps!

Best,

Kerstin

[Stan Jones]

Thanks, Kerstin. That confirms how off base my workarounds were!

This is what I see:

A subclip created from a source file with a transcript keeps the transcript (with any edits to the transcript), and the timecodes adjust in the timeline based on its placement in the sequence. That is as expected.

At first, I thought that the transcript does NOT keep any changes to the Speaker Names that are made before subclip creation. Double-clicking the subclip in the timeline and then editing its transcript speaker name produced odd results: once it correctly showed the right Speaker/clip. another time, it showed the speaker name for subclip 1 for subclip 2 also (which had a different speaker name).

If you edit the source clip transcript AFTER creating the subclip, the subclip transcript does NOT change. That is what I would expect.

An edit to subclip 2 showed when subclip 2 was opened in the source monitor. It did not change the original source clip. Both, are what I would expect.

@gabrielk41156439,

I did not say it before, but this is a complex workflow. Tracking dialogue in a group is no small undertaking!

> when working on fixing a transcription from a sequence which includes multiple microphone channels from separate recorders , which in this case I used to capture a large group conversation, I don't know which source file I'm working with, meaning that I edit the transcript on the sequence but dont know if the changes are being copied to every source file on the timeline or which one is being actually rewritten aside from the sequence transcript.

I promised to come back to this. Given the problems with subclips, I don't have a specific recommendation. But here are 2 things to consider.

1 - I have experimented with using Speaker names to identify the transcript sections of multiple clips in a sequence. The names must be edited in source view. Most often, each speaker will be in a single clip. So rename "Speaker 1" to "Mary," and when you add the clip to a sequence, any part of the transcript that is from Mary shows that name. With multiple speakers in a single clip, and multiple clips that contain a particular speaker, it becomes more complex. But I suspect that is what you are looking for.

Are I noted above, if you are using subclips, the Speaker name editing must be done after the subclip is created. But I got inconsistent results.

2 - Staff have reported that "when you have multple transcripts of multiple clips stacked on multiple A/V tracks, we only show the transcript of the highest ranking audio track (usually A1). You'd have to mute the higher audio track if you want to see the transcript on the lower audio track." And this is what I see. If you have audio tracks overlapping, you only see the transcript of the highest, unmuted track, that has a transcript.

Stan

[mafaldarebelo]

I need some help as well. 
I'm editing a multi-cam and multi-mic sequence of an interview in portuguese. It's all in sync in one first sequence (2 video tracks and 5 audio tracks) 
I went to the text-based-edit panel and made a transcription of only one file (chose the best audio and used it to serve as guide for the transcription) in Portuguese. 
The transcription was made in the sequence. 
The transcription was not that great and so I've used the feature "export to txt" and "import corected transcript" to replace it with a good transcription. So far so good. The sequence reinterpreted well and I was able to cut and move the text influencing the sequence files (all of them). 
I then made a copy of the sequence to start editing. The transcription seemed to transfer as well. I've tested copying a chunck of one timeline into another and changing text order and the transcription was going along and working well. So I proceeded to edit the interview by selecting in the TBE panel and cutting the timeline, not deleting anything. By the end of this selection process, I tried to delete all the unwanted parts and it's greyed out. It seams that the transcript became static and related to the timeline times and not to the video content, although it's not assuming to be static. When I copy a portion of the timeline, the whole transcript comes along without text-to-time connection. 
It's really important to keep the transcript as a TBE in this process and be able to rely on that.

Any intel of what might have happened and how to get back to a funcioning TBE? 
I've already selected 7h of interviews. 

[Stan Jones]

@mafaldarebelo,

Thanks for that excellent description of your workflow. You are in the Beta forum; what version of PR are you using? And if the Beta version, what Build?

Here is what happens with timecodes and transcripts. The Source media transcripts are always the whole, original clip/file on disk. The timecodes in it are the timecodes from that source. A "static" transcript of a sequence is always what was transcribed at the time you made the transcript. (Unless it is nested, and that is a different issue.) Those timecodes do not change as you edit the sequence. The "sequence transcript from source clips" is what is left in the sequence with edits you have made, and the timecodes in the transcript are not from the source clips, they are from the sequence.

But in all instances, the audio should match the transcript.

> transcript comes along without text-to-time connection. 

Are you referring to the audio matching the transcript? Or timecodes?

The details matter here, so I'll respond further after your response.

Stan

[Stan Jones]

Stan

[mafaldarebelo]

Hello:

Thanks for the quick reply. 

I've edited most of it in PP25.1. At some point, reading forums, I've installed and tried Beta version 25.3, with no better luck in this transcription workflow. 

> transcript comes along without text-to-time connection. 

Are you referring to the audio matching the transcript? Or timecodes?

I copy a chunck of the sequence into another 

I'm not sure how to proper explain everything because I even tried a few tests that have worked in a moment and not in another. Sometimes when I shut down and reopen the app, the transcription behaviour changes as well.

So I'm looking for the right TBE workflow for this:

 There are several clips in several tracks. 

I felt it didn't make sense to transcribe all files in there, because they're all the same interview and I would have to "import corrected version" of the transcription for all of them (the Premiere portuguese transcription is poor, so I needed to edit a lot of it, replacing it with a better one (like TurboScribe)).

[It would be great to have a feature in which we could feed the project with the proper transcript and it would do the recognition from that source.]

But right now I feel like the whole process is a mess. 
I'm willing to go back and start over, if there's guaratees that with a proper workflow it might work. 
Because I'll be doing a lot more interviews and it has to be reliable. 

With these needs, what would be the correct approach and workflow?:

- several tracks (video and audio) in sync

- needing to replace the transcription with a proper one.

- willing to have full interviews (as synced sequences) transcribed to then start intercalating them in another sequence (having the related transcriptions in place). 

Thanks,
Mafalda

[Stan Jones]

Mafalda,

Thanks for the response. That is helpful. I do understand how explaining your workflow is difficult, particularly since what you see can change.

I like your workflow of using one audio source for the transcription. Is that also the source for the audio to be used in the final export?

I asked about your statement that the "transcript comes along without text-to-time connection." Your description that you are copying from one sequence to another is helpful. But I still need to understand what you think is incorrect about what you are seeing. Is there an audio sync problem? In other words, does what you hear not match the transcript? Or does the timecode in the transcript not match what you are expecting?

Stan

[mafaldarebelo]

>>Is that also the source for the audio to be used in the final export?

The audio used for the transcript will be in the final sequence for the export, yes!

>>I asked about your statement that the "transcript comes along without text-to-time connection." Your description that you are copying from one sequence to another is helpful. But I still need to understand what you think is incorrect about what you are seeing. Is there an audio sync problem? In other words, does what you hear not match the transcript? Or does the timecode in the transcript not match what you are expecting?

Up until a point, the transcript is connnected. That means that if I copy and paste a part of the transcripted sequence, only that part's transcription comes along still recognising the playhead and playing properly. 

On the last phase, this connection has been lost. If I copy a part of the sequence from 15:00:00 ultil 16:00:00 and put in another sequence at the timecode 09:00:00, the whole transcript comes along and when I hit play, the transcription displayed at 09:00:00 is disconnected and unrelated, and the proper trasncription is at timecode15:00:00. 

I hope I've made myself clear. 

[Stan Jones]

@mafaldarebelo,

Yes, that clarification is very helpful.

Are you using any subclips or linking any audio with video? I think it was subclips that produced inconsistent results with exporting/copy/pasting or similar. 

I'll try to test the workflow you are describing.

Stan

[mafaldarebelo]

No.

I only used a color label for the whole sequence. And have cut the clip, selecting the parts I'm interested in, by moving the top video one track above. 

thanks for trying to help. 

[mafaldarebelo]

There's no other clue on how to proceed? 

Thanks