• Global community
    • Language:
      • Deutsch
      • English
      • Español
      • Français
      • Português
  • 日本語コミュニティ
    Dedicated community for Japanese speakers
  • 한국 커뮤니티
    Dedicated community for Korean speakers
Exit
1

Feature request: Bilingual transcripts

Explorer ,
Apr 18, 2024 Apr 18, 2024

Copy link to clipboard

Copied

As a Canadian editor who works with English and French, often in the same project, it would really be useful for me to be able to produce bilingual transcripts.

 

At the moment, when I transcribe a bilingual interview, one of the language comes out as gibberish. I need to do two passes of transcripting, export as CSV and merge in Excel. It's very impractical.

 

Premiere is able to transcribe both languages reasonably well, the next step would be to auto-detect both in the same transcript.

 

Thanks!

Idea No status
TOPICS
Import and ingest , Sound , User experience or interface

Views

244

Translate

Translate

Report

Report
Community guidelines
Be kind and respectful, give credit to the original source of content, and search for duplicates before posting. Learn more
community guidelines
6 Comments
Explorer ,
Apr 18, 2024 Apr 18, 2024

Copy link to clipboard

Copied

I faced this too, my interviews were in Hindi and English. And when I transcribed for English, you guessed it, the Hindi bits were gibberish, and the other way round.

 

I found a solution, I think. You can select in and out points in to select the English bits in your timeline and select the transcription settings according to your language. Then you can set your in and out points to select the French bits, or the other language and check the "merge output with existing transcription" and keep adding to your original transcript.

Votes

Translate

Translate

Report

Report
Community Expert ,
Apr 19, 2024 Apr 19, 2024

Copy link to clipboard

Copied

Upvoted. But I suspect this is not as simple as we would hope. As opposed to identifying Speakers, they need to identify what language is spoken and then apply that language model. But a bilingual speaker may switch every few words. How do you deal with that?

 

@Yash Nene, I like that workaround as something that works, even if it is potentially a lot of work.

 

A workaround may be coming for the situation where one speaker only speaks one language: in the Beta, you can filter by speaker and bulk delete. So transcribe, select the non-English speaker and delete (without removing space in the timeline); repeat for the other speaker. The trick is whether they can be combined or only at a captions level.

 

Stan

 

Votes

Translate

Translate

Report

Report
Explorer ,
Apr 19, 2024 Apr 19, 2024

Copy link to clipboard

Copied

Thanks Stan for the upvote. I realize it may not be easy.

In my case, the speakers did not change language every few words. They
spoke consistently English for a couple of minutes, than another person
would speak French for a few minutes.

In any case, perfection would not be necessary for me. I could handle a few
mis-transcriptions. 80% recognition would be a success.

That may help make it a bit easier, perhaps?

Sylvain

Votes

Translate

Translate

Report

Report
Explorer ,
Apr 19, 2024 Apr 19, 2024

Copy link to clipboard

Copied

Thanks Yash Nene for the tip. This could help as a workaround!

I did not know you could transcribe just part of a clip using In/Out
points. Or merge the results.

Could you explain your workflow a bit more?

Thanks very much!

Sylvain

Votes

Translate

Translate

Report

Report
Community Expert ,
Apr 20, 2024 Apr 20, 2024

Copy link to clipboard

Copied

Note that transcribing in to out only works with STATIC/Sequence transcriptions. This is true even though you can place in/out markers in the Source Monitor that could be applied to a source media transcription.

 

Generally, set an in-point for the beginning and an out-point for the end of the section you want. It helps to set them where there is no speech. In the "Generate Static Transcript" dialogue, expand the Transcription Preferences arrow if it is collapsed. Check "Transcribe in point to out point only." Check the box for "Merge output with existing transcription." Click the Transcribe button.

 

  • If there is no existing transcript, the box for "Merge output with existing transcription" is greyed out.
  • If you do not set an in/out point, the option for transcribing in to out is greyed out.
  • If you do not set BOTH an in and out point, or if your out point is "out of bounds," you can check the "in to out" box, but the Transcribe button is greyed out. (See this post for more information: https://community.adobe.com/t5/premiere-pro-discussions/transcribe-quot-in-point-to-out-point-only-q...
  • If you check the "Merge output with existing transcription," it keeps any existing transcription that is not in the in-to-out range, with any edits you have made to those sections.
  • If you do NOT check the "Merge output with existing transcription," it deletes the entire sequence transcription, and replaces it with ONLY the in to out section.

 

Stan

 

 

Votes

Translate

Translate

Report

Report
Explorer ,
Apr 20, 2024 Apr 20, 2024

Copy link to clipboard

Copied

LATEST
Thank you Stan for this detailed workflow. I can't wait to try it out!

Sylvain

Votes

Translate

Translate

Report

Report