Skip to main content
TeresaDemel
Adobe Employee
Adobe Employee
August 2, 2022
Question

Text Panel Enhancements (Part 2): Assemble Rough Cuts from Transcripts

  • August 2, 2022
  • 36 replies
  • 30197 views

As you saw in my earlier post about text panel enhancements, transcripts are becoming more central in editing workflows. Starting today in Premiere Pro (Beta) you can create rough cut edits from selected Text passages.

 

To start, you can generate transcripts of your footage in the background when you import your media. Alternatively, you can generate transcripts from the Text panel after your project has been created. Either way, the Sequence transcript panel dynamically syncs with your Timeline: as you make changes in your Sequence transcript, the Timeline will reflect your changes – and vice versa.

 

My colleagues and I created some starter footage for you, as well as a tutorial so you can try it out. If you tend to do all of your editing in a Sequence, follow the guidelines below and/or watch this tutorial. If you tend to do three-point (Source) editing, follow the guidelines below and/or watch this tutorial.

 

Sequence Editing Workflow

  • Download the four videos from the Five Questions video assets folder
  • Open Premiere Pro (Beta -23.0x006 or higher) and create a new project
  • In the Import screen
    • Import "Kerstin interview"
    • Turn on Automatic Transcription and make sure “Create new sequence” is checked. (Note: You can turn on and off automatic transcription in Preferences -> Transcription -> Automatically transcribe clips on import.)
    • Make sure language is set to English
    • Click "Create"
  • In Edit view, open the Progress Panel (Window -> Progress). This will allow you to see transcription progress.
  • The clip should be in your new sequence
  • Click on the Transcript tab in the Text Panel
  • Click “Transcribe sources.” This setting allows your Timeline and Transcript to stay in sync.
  • Use the Search box in the Transcript panel for the word "what." This will allow you to see where the questions are in the video 
  • For the first question, highlight everything from "I had a piece of that cake a while ago" until she says "Second question."  
  • Click the In/Out point at the top of the Transcript panel. You should see that section of the video highlighted with In/Out points in the Program Monitor
  • Delete the highlighted section using the "Extract" button in the Program Monitor.
  • Go through the rest of the transcript and cut it down to a simple Q&A highlights reel. Have Kerstin state the question and a simple answer, but delete extraneous information. For example, have her say her favorite karaoke song but exclude detail on what makes it a great karaoke song.  
    • If you wish to have Premiere automatically set In/Out points around any selected text, go to the dropdown menu in the text panel and select "Set In/Out around selections."
  • Play back your sequence. You should have one sequence with five questions and simple answers.
  • Play back your clip and check for any issues.
  • Try it with your own projects!

 

Three Point (Source) Editing Workflow

  • Download the four interview videos from the Five Questions video assets folder
  • Open Premiere Pro (Beta - 23.0x006 or higher) and import the four videos to your Desktop or other preferred folder
  • Create a new project 
  • In the Import screen
    • Check the boxes next to all four videos
    • Turn on Automatic transcription (Note: You can turn on and off automatic transcription in Preferences -> Transcription -> Automatically transcribe clips on import.)
    • Make sure “Create new sequence” is unchecked
    • Make sure language is set to English
    • Click "Create"
  • In Edit view, open the Progress Panel (Window -> Progress). This will allow you to see transcription progress.
    • Pro Tip: Set up your windows so that Text Panel, Source Monitor, and Program Monitor are side by side.
  • Create a new sequence (File-> New -> Sequence). Name the sequence "Karaoke"
  • Double-click on any clip in the Properties Panel
  • Click on the Transcript tab in the Text Panel
  • Click on the “Source” tab to keep the source transcript in view
  • Search the transcript for the word “karaoke.”
  • Select the question and answer and click the In/Out point at the top of the Transcript panel. You should see that section of the video highlighted in the Source Monitor
  • Use the Insert button in the Source Monitor (or comma (,) keyboard shortcut) to insert the selection into the timeline.
  • Do the same thing for the four remaining source clips
    • If you wish to have Premiere automatically set In/Out points around any selected text, go to the dropdown menu in the text panel and select "Set In/Out around selections."
  • Play back your sequence. You should have one sequence with five people answering the same question.
  • Switch to Sequence view in the Transcript tab
  • Click “Transcribe sources.” This setting allows your Timeline and Transcript to stay in sync.
  • If there are any extraneous words in your sequence, highlight those words and click the In/Out point at the top of the Transcript panel. You should see in/out points selected in your Timeline panel.
  • Delete the words and associated subclip with the Extract (apostrophe (‘) keyboard shortcut in the Program Monitor.
  • Play back your clip and check for any issues.
  • Try it with your own projects!

 

Stay tuned, because we have more coming this year

  • Document style text editing: Use Cmd+X/Cmd+V and Delete to edit your tracks from the Text panel.
  • Range selection: Select a sub-clip on a single track using text. We know that In/Out points + ripple delete affects all tracks in your sequence and deletion isn’t helpful if you have B-roll, music, etc.
  • Language detection: Set your language preferences in Preferences -> Transcription for now. The ability to detect language on import will reduce a click, and can be helpful when you are working on a project that includes multiple languages.
  • Search Source footage for key moments

 

We would love to hear from you

  • How is the performance of automatic transcription when you have hours of files? Please share your system specs + the number of hours of source footage you are transcribing.
  • When working with your own footage, are you mostly editing the Sequence Transcript, Source Transcript, or a mix of both?
  • Are there any moments where the UI or descriptions feel unintuitive?
  • How well do the timing of the cuts work?
  • For non-English languages: Are there any languages or dialect problems, e.g. problems selecting text, editing word boundaries, etc?

36 replies

Known Participant
November 20, 2022

Hi there.  I am trying this workflow with clips that have audio and video recorded separately.  If I import the source audio and video with automatic transcription turned on, then merge clip is greyed out. And there's no way to merge the clips unless I create a new project with automatic transcription turned off.  If I do that (turn off automatic transcription) and import the clips, I can merge them.  But then if I open the text window and hit transcribe source, I get a progress bar in the text panel, but it never turns up in the progress window.  And it never completes.

 

The only work around I've found is to EXPORT the merged clip as a quicktime, and then re-import it in a new project with automatic transcription on. Inefficient, but a workaround.

 

So I guess what I'm understanding is, this feature ONLY works when audio is recorded in the same file as video? And negates the workflow Premiere has implemented for people using DSLRs and recording audio and video separately, then using merge clip to combine them in post?

 

Or did I miss something here???

 

It's a great feature.  But I am teaching a course where all our students use DSLRs and send audio to Tascam recorders.  I was planning to teach this new feature as a way for them to simplify their workflow, but suddenly it's made their workflow more complicated.

Kerstin Ebert
Adobe Employee
Adobe Employee
November 22, 2022

Hello,

thank you for reaching out and reporting this problem! This is very valuable feedback and we are working on a fix for this so that you will be able to work with source transcripts for merged clips!

 

Best,

Kerstin

Known Participant
November 3, 2022

What really needs to happen is for Premiere to be able to upload a standard txt file of the text, and then put it into a transcript that syncs with the audio. This is needed if the speaker is using a teleprompter, and the text file with all the spoken words already exists and the speaker is simply reading it. To use the auto-transcription method produces a huge number of errors that have to be corrected manually, and that is the most time consuming part of the transcription process. But if one already has a text file with all the words in it, you shouldn't need to use auto-transcription in the first place. YouTube allows you to do this with uploads. You just upload or cut and paste the raw text as one big block, and it slices and dices and sorts out the time codes. Anyone using a teleprompter needs this capability in Premiere.

Celestial_days0D45
Participating Frequently
December 13, 2022

Agreed. I don't like to have to fix captions when I already have a script I read from.

Participating Frequently
November 3, 2022

Hey - first, I just wanted to say thanks SO much for developing this. 

As far as feedback, indeed it feels a bit clunky but I assume these are growing pains. 

- It should be possible to select text through breaks in the transcript.

- I hope this doesn't slow down editing?

- Really just take a look at Descript's interface, they've been doing this for a while

- For example, um/ah detection/removal, coughs, breaks, spaces

Thanks again!

Participant
October 11, 2022

This really is a great feature! When you say "Search Source footage for key moments" is coming later this year, is this a global search function? It would be very useful to be able to search across all your source footage. 

Participant
October 11, 2022

Also, the new "Whisper" open source ASR from OpenAI gives me significantly better results in Norwegian than the model Adobe uses. Any plans to implement Whisper in your products? 

Thanks!

Stan Jones
Community Expert
Community Expert
October 5, 2022

@TeresaDemel I finally started playing with this, but have not tested enough to understand the issues.

 

Here are two threads in the regular forum that you may want to look at regarding users' workflow interests. Does the new workflow work if you use the source-sequences in multicam? Can a transcript-as-srt, designed for captions, be modified as needed for format and imported as a transcript and used with this workflow? Sorry I have not  been able to experiment with this more.

 

https://community.adobe.com/t5/premiere-pro-discussions/trying-to-edit-multicam-clips-with-subtitles-into-timeline-and-keeping-multicam-functionality/td-p/13245187

 

https://community.adobe.com/t5/premiere-pro-discussions/srt-s-not-editing-from-a-multicam-sequence-into-a-timeline/m-p/13243712#M433550

 

Stan

 

Participant
September 20, 2022

Will I be able to use the transcript in other application like Prelude?

Our journalists are still using Prelude for watching and organising footage. Or even to do transcription manually. So is the transcipt part of the metadata of a clip?

Community Expert
January 19, 2023

@detlefs23288115 

I just noticed that no one replied to you.

Prelude is discontinued.  (Unfortunately, I'd say.  It was really great having something slanted toward reviewing, logging, and transcribing footage without everything else we find in Premiere Pro.)

It's looking like rather than manually transcribe, we'll do Automatic Transcription and then correct the resulting transcript in the Transcript tab of the Text panel.

Kerstin Ebert
Adobe Employee
Adobe Employee
September 5, 2022

Hi Joseph,

can you please walk me through the steps you followed when running into this bug? Were you editing text in a source clip transcript or in a sequence transcript? In which beta build are you seeing this?

 

Thanks,

Kerstin

Stan Jones
Community Expert
Community Expert
September 7, 2022

Hi, Joseph,

 

I'm sure you're most wondering if @Kerstin Ebert got a look at it!

 

I've been out of town and unable to experiment further. I do see the problem in your video, but I cannot replicate it on two different machines. Since you indicated it has been present for some time, I tested in 22.5, not the Beta. Both machines are Win 10, however. I created some text layers and edited text - with regular and then aligned right.

 

I have seen similar glitchs occassionally, but it is momentary, immediately corrects, and does not freeze. I think the last I saw it, I was changing scaling and position on 4K GoPro clips. It was not with text.

 

Stan

 

Known Participant
August 29, 2022

This looks fantastic!  I'll report back after testing it out.

DJP2014
Inspiring
August 23, 2022

Really love you're including this in Premiere. My feedback is specifically targeted at 3 Point Source Editing.

 

It seems a shame now we have 'Set In/Out around selections' that we have to leave the Text Panel to edit to the timeline. Working through a long transcript would be so much quicker if we didn't need to keep switching to the Source Window. Instead, one hand would make text selections and the other keyboard commands for insert/overwrite.

 

Another thing it'd be great to see is a way to speed up stringouts. Was thinking you could include a special command like an ellipsis in the Search panel. Something to instruct Pr to select the whole of a question - i.e. "Start of question ... end of question".

 

With every take in the transcript highlighted an additional command (Edit Multiple selections?) could be added which would edit every take into the timeline. Stringouts made simple!

TeresaDemel
Adobe Employee
Adobe Employee
August 24, 2022

Thanks, DJP! 

If I understand correctly, would you want to do the following? 

  1. Open up transcript for Source Clip #1
  2. Select dialog for a take (or part of a take)
  3. Insert dialog /take into sequence
  4. Select matching takes in Source Clip #2, Source Clip #3, etc using Sensei magic
  5. Insert all selections into sequence. Instant stringout!
DJP2014
Inspiring
August 25, 2022

Not quite, see if this makes sense...

 

1. Open up transcript for Source Clip #1

2. Search dialog for the whole take using new elipsis functionality. In this example the search query would be: "My name... Lynch". Current behaviour already highlights every take, no Sensei magic required.

3. New command would insert 3x highlighted selections into sequence. Instant stringout!

 

 

When the take i'm interested in spans multiple clips I'd simply nest them and use that as the Source Clip in stage 1. *Making sure before inserting them into the sequence that 'individual clips' was selected in the timeline to avoid inserting the nest instead of the clip.

 

 

Adobe Employee
August 11, 2022

Great progress team! I am a long time Premiere user (started using it to make conference videos 20 years ago), but I don't use it often (maybe twice a year or so).  For transparency, I work at Adobe, so you can reach out to me directly for debuging.

 

Here is my feedback:
  • I didn’t see the transcription option as part of import. I'm not sure I looked at any of the option in the panel on the right.
  • Why did I have to transcribe twice - once on import and again in the Text panel? It seems like I should just be able to do it once.
  • I imported two files - Mary talking and Kerstin talking and tried to interleave their answers so that each would answer the same question before moving on to the next one. So I had two clips in my sequence - one of Mary and one of Kerstin. I wanted to interleave their introductions to start.
  • I followed the instructions in this webpage. I couldn't follow this instruction: Delete the highlighted section using the “Extract” button in the Program Monitor. I wasn't sure where this button was located. So I used Apple help to search for it and found it as a menu. When I clicked Extract the video was edited but the text interface didn’t reflect that I had cut out some content. I think this must be a bug. 
  • I actually didn't notice that I had run into a bug at first and went on to find other pieces but then because the text didn’t reflect my sequence any longer I could no longer use it to select the right parts and gave up. 
  • I also got a crashing bug on import and tried to submit it through the file this bug button but then I couldn't log into Jira and so I didn't file it.
TeresaDemel
Adobe Employee
Adobe Employee
August 12, 2022

Thanks, Mira for trying this out! You bring up a great question in bullet #2 and I would love for others to share their preferences here.

 

Why did I have to transcribe twice - once on import and again in the Text panel? It seems like I should just be able to do it once.

 

There are two transcription modes for a sequence:

  1. Transcribe all source footage and have the source transcript sync with the timeline transcript.
  2. Transcribe the sequence and "break" the relationship between the source and sequence. 

 

The second type of transcription is what happens in our existing system (July '21 - now), where users would transcribe at the end of their workflow in order to caption the final video. Potentially users want to have a "fixed" transcript in order to have some assurance that the transcript won't change. 

 

However, the first type of transcription will still allow for captioning. 

 

I would love to hear from the crowd here: Can we eliminate the second option or are there use cases where having a choice is important? 

 

Regarding the discoverability of "Extract" (and I assume "Insert") -> Those issues will likely be remedied once we have more Word doc-like keyboard shortcuts. Your comment about those being unfamiliar and difficult to find is representative of users that don't tend to hang out in our Beta forum! So thanks!

 

Inspiring
August 22, 2022

Hi Teresa

 

this is an awesome feature and I am about to dive in to test it. It is one of the holy grails. 

I have a very large production/ project with maybe 80 interviews thst I have distributed across multiple projects within one production and for which I have already cut "sync pulls" or "sync selects" with the intention of using those sequences as my master sources, probably using pancake editing. These sequences are already transcribed (initially last year using the old method) and act as an intermediate stage in my workflow. 

I am hoping that I am going to be able to use text from those sequence transcriptions as my source text for making a rough cut/paper edit but as they are across multiple sequences wonder if this is going to be easy?

Whatever the answer/outcome to this question, I would say keep the second, older method anway.