Adobe Premiere transcription problems/improvement suggestion:
(This following is all on good quality dialogue using good mics, clear British accents, no background noise, etc).
1. Biggest problem that takes a lot of time to correct - punctuation is poor, it often puts a full-stop in unnatural places (that should be a comma, for a sub-clause in a sentence), but then it does not capitalise the first word in the following sentence (whether correctly or incorrectly a new sentence).
This seems a simple coding fix..? If This Then That. Then I wouldn’t need to go in and correct as often.
Many times in spoken language, a comma, or a new sentence doesn't matter so much for understanding the dialogue. but a full stop then no capitalisation breaks the flow and often creates ambiguity in meaning. sometimes requiring a second read of a sentence.
(See what I did there).
Also, my clients often want grammatically correct captions/transcriptions! So I gotta change it… Sometimes captions are just an aid to the audio, but some people rely entirely on the captions/transcript.
2. The editing transcription process is a little cumbersome - you need to double click on a word in a paragraph to make that whole paragraph editable (if you’re in a hurry and miss, instead double-clicking on a space, it doesn’t open), then press escape to save the change and go out so you can continue to Play and review. It’s just more clicks and time. I’m not sure the exact solution (or am I missing a function?), but something to work on for future versions.
3. I have language set to British English but it still uses American spelling in many words, eg “z” instead of “s”. Need more options in Preferences.
4. Then there is general mis-hearing/mistaking words, understandable this will happen at times, but it is less accurate than the best AI out there, hoping Adobe’s transcription will improve! Or integrate OpenAI’s Whisper? I’d love the option of using remote/cloud large model for better accuracy. (I am separately looking into using Whisper then importing .srt, but there are pros/cons to that).
I quite understand technical language may not be recognised and accept correcting that (eg I film/edit/caption a lot of medical conferences/interviews).
I tested the transcription with a friend's much newer PC, compared to my 10-year old iMac - no improvement in accuracy on the above problems. Speed isn’t the issue in this case - even my old iMac creates the transcriptions perfectly fast enough - accuracy is the important issue for me.
Other than this, editing from transcriptions is a great help in Premiere (searching phrases/words, jumping to places in the video etc), just need the accuracy to improve!