[ Updated Nov. 7, 2021, with significant changes to text and screen shots throughout, especially regarding transcript exports. ]
Adobe released a new transcript/caption workflow with the July, 2021, release of Premiere Pro, then significantly updated it in October, 2021, with the v22 release. While captions were a part of Premiere for a while now, the automatic speech-to-text transcription is new. Here’s an overview of how this works.
BEFORE YOU START
While you can create transcripts at any time, there are two places where transcripts will be the most helpful:
Let’s take a look at both. While the initial release had problems with text exporting, Adobe fixed them in the v22 release of Premiere Pro.
TRANSCRIBE BEFORE YOU START
Both transcripts and captions are created from clips in the Timeline.
Edit the clip, or clips, you want to transcribe into the Timeline. Make sure the audio volume in the clip is loud enough to hear. And, if there are tracks that just contain sound effects or music, mute them.
NOTE: As part of the transcription process, if you assign tags to your clips – like Dialogue, Effects, Music… you can specify that Premiere just create transcripts of Dialogue clips. This means you don’t need to mute other tracks.
NOTE: To mute a track, click the M icon in the track header for that track.
Select the timeline, so there is a blue box around it, then switch to the Text panel and click the Transcript button in the top left. (Whether you click the Transcript or Captions button, you’ll still see the same three choices.)
NOTE: The Text panel is new. There used to be a Caption window in its place, which is now gone.
Click the Transcribe sequence button to start the transcript process.
The Create Transcript window appears. Most of the time, the defaults are fine but, if you are working with a language other than English, be sure to set the Language menu correctly.
Then click the blue Transcribe button to start the process.
Premiere extracts the audio from the sequence and sends it to the Cloud for transcription. As well, Premiere displays screens indicating the transcription process.
After a bit, which varies depending upon the length of the clip, the transcript appears.
To change the name of the default speaker, click *** Unknown and select Edit Speakers.
NOTE: Don’t add a new speaker, otherwise you’ll need to change all the existing speaker names manually.
To edit a name, click the pencil and type the new name.
To add an additional speaker, click the Add Speaker button.
With the v22 release. Adobe changed the behavior of this window. It now expands as you add new speakers, making them easier to see. To edit a speaker, click the pencil. To delete, click the X.
When done, click Save.
To change the speaker associated with a paragraph of the transcript, click the speaker name and pick the correct name from the list.
To edit a transcript, say to change a word, correct a spelling or add punctuation, click the paragraph you want to change and, like any text editor, edit it.
NOTE: Currently, Adobe tends to put the period at the end of a sentence at the start to the next paragraph. Keep your eyes open for this. It is easy to fix with manual editing.
When the time comes to do something with that transcript, click the three dots at the top right of the Text window to reveal the export menu. The v22 release now provides three export choices:
All three export options display a Save dialog so you can name the file and give it a storage location.
Export Transcript is used to create a .prtranscript file that can be opened in the Transcript panel using the Import transcript option. Use this when moving the transcript from one Premiere system to another.
NOTE: Use Import Transcript from this same menu to import a transcript that was exported from another Premiere system.
Export to Text File. Use this option to create a .txt file for proof-reading, to share with a client, or to create written content for your video.
NOTE: Select Display pauses as […] to display pauses as ellipses so that the transcript shows where there are gaps in the dialog. This is very useful to know where you’ll probably need B-roll.
The big news with the v22 update is that exporting a transcript as a text file now includes both the timecode reference and the speaker’s name. This makes these transcripts perfect for client review and editing your talking heads on paper before you start moving clips. This was a much needed improvement from the initial release.
Another new feature in the v22 update is the ability to export transcripts as CSV files. This puts every element into it’s own spreadsheet cell. These can be opened in Excel, Numbers, or any program that accepts CSV files. This is another very useful feature.
Another beneficial v22 change is that you can search for text across all clips or all captions in a transcribed sequence. This is because the entire sequence transcript is loaded into the Text window when a sequence is opened. This makes finding a specific clip containing specific text very fast.
NOTE: If you add clips to a transcribed sequence, select the timeline, then choose Re-transcribe Sequence from the three dot menu at the top right corner of the Text window.
The second major use of transcripts is in creating captions. The first part of the process – creating and editing the text itself is the same.
Once a transcript is created and edited, coverting into captions is amazingly easy. Click the Create Captions button.
The Create Captions dialog appears. Most of the time, I just accept the defaults, but you are welcome to tweak.
Several things happen at once:
At this point, adjusting and trimming captions is the same as earlier versions of Premiere.
NOTE: When trimming captions, always use the red Roll tool. Trimming with a yellow Ripple tool will knock all downstream captions out of sync.
The improvements to text transcripts make this feature seriously useful. As well, the new transcript/caption workflow is fast and easy. Best of all, Adobe is building the cost of transcription into your monthly subscription fee; at least for now.
If you haven’t looked at the new options for Speech-to-Text in Adobe Premiere Pro, now is a great time to start. This one feature can save you hours figuring out how to edit a project – especially one filled with interviews.
19 Responses to New Transcript/Caption Workflow in Premiere Pro [u]
I was so excited to try the new speech-to-text transcription, and so disappointed when I opened the txt.
file and saw that there was no timecode or speaker references. A glaring omission, indeed. On the bright side, I compared the Adobe transcription to a Sonix transcription of the same interview and can report that the Adobe transcription was slightly better in a few places with word recognition and slightly less accurate in some places with speaker recognition; par for the course with cloud transcription comparisons. I’m crossing my fingers that Adobe fixes the timecode and speaker issue in a timely fashion. It would be very cool if they eventually introduced some of the more advanced editing features that Sonix provides, but you can’t expect everything for free 🙂
Larry, I wish you an easy move and hope you settle in well in your new location.
Thanks for your comments and kind words.
Yes, I was REALLY surprised that, for a feature that went through such a long beta, the inability to export timecode and speakers was not corrected sooner.
This is a nice feature, but this needs to get fixed.
Good walk-through, Larry. You saved us some serious wasted time.
The speech-to-text is a great feature, almost. The omission of TC makes this useless for our rough cuts, as you mentioned. I’m hoping Adobe will soon add the ability to select “Export to text file with TC” in the export window.
Until then, we will not be using this feature very much. Transcription services such as 3Play are great for captions, allow editing within their interface, and are quite inexpensive for shorter projects.
Yup. That’s why I featured this omission in my tutorial. Hopefully Adobe fixes it soon.
Does the captioning give punctuation and capitalization, or do I have to go back thru the entire dang thing and manually fix every consarned sentence? If no, then this is hardly an improvement over bumming free captioning off youtube, unless it’s considerably faster. About how long did sensei take to transcribe/caption a half hour of speech?
No service does punctuation correctly – Adobe is no different from anyone else. So, yup, proof-reading is essential. The big differences to Adobe vs. YouTube are:
* Tight integration into Premiere
* Higher accuracy – YouTube is notorious for bloopers
* Instant convertibility into captions
I’m testing the transcription feature on a recent interview. Results will be back in about an hour. I’m curious about security in the cloud though. Who has access to the transcript besides me?
Good question – I have no idea how Adobe is handling security for transcripts. That would be worth asking them.
There is a workaround for the missing TC from the text file export of the captions, though it is a bit kludgey. If you create captions from the transcription, you can then export the captions as an .srt file. If you edit the captions before exporting the .srt, you can combine the two-line default clips into larger paragraphs. Which essentially is what the transcription text export should have been. The editing of the captions into essentially a timecoded transcript is just a bit time-consuming.
Kludge is the right word. Thanks for the work-around, BUT…. Adobe needs to put this into Premiere itself.
Can PP import a txt file with the speakers transcript that does not have timecode,
analyse it – ideally offline, and create captions (closed or open) at the right timings & export to srt?
No. Captions require timecode. There’s no way to sync a text file with project audio to create captions unless timecode is part of the import.
I found the “kludge” method too with .srt files but the timecode notations happen at intervals so short that reading the transcript is hard. Hopefully export to text WITH timecode will happen soon.
One work around for the Timecode missing with Transcription… And that is to convert and use the Captions. Merge what you want together and export an SRT. Open in a text editor, add the speaker names and save it off
The only problem with this approach is that if you are handling a client two hours of rough footage to help find selects, all that timecode gets in the way of reading the transcript. And getting rid of it is not easy.
If you open the SRT file that premiere exports in another subtitle program like Annotation Transcriber you can output an RTF with time-code file that at least gives you the TC for reference for review purposes. This is also handy for creating multiple language translations for a wider subtitle coverage.
Ideally Premiere should build in support for RTF with time-code.
Thanks for your comments.
Larry, Is it possible to export the final transcript (after edits) so that a transcription person can clean up the edits, and then REIMPORT the transcript into Premier?
ONLY if you export it using the Export Transcript function. This requires that the transcriber correct the transcript in their version of Premiere. It can’t be corrected outside of Premiere.
You can not export a transcript as a text or CSV file, correct it, then reimport that corrected file. Text and CSV files don’t have sufficient timecode references.