This content has been marked as final. Show 7 replies
That is to be expected. Even dedicated speech recognition software - after extensive training - has difficulty.
I've got good (not perfect) results from the speech to text function, but only when it was a speaker who was properly mic'd up, spoke clearly, and spoke at a moderate pace.
"That is to be expected. Even dedicated speech recognition software - after extensive training - has difficulty."
Exactly. Waste of programming time in my opinion. Maybe it would be useful for talking head interviews and the metadata would be useful for huge post houses. Except that huge post houses don't use Premiere.
Though I am rooting for Adobe to get their head out the sand and make 'Premiere' truly 'pro'.
It's the first step in something larger. And I can say with confidence that once it gets more developed it will be a huge help in workflow, even at the lower end. I've met more than a few producers/production managers who are interested in this functionality.
Yeah, once speech recognition technology works out of the box with out training we can get rid of keyboards. Problem is, it's not there and there is a way bigger market for speech recognition than NLE use. I just don't think Adobe should waste resources on it yet.
There a ton of other ways Premiere is lacking, but Adobe seems to think you have to add new features at the expense of having a solid program. Premiere CS4 before 9.0.1 update was so bad I was embarrassed for Adobe.
"I've met more than a few producers/production managers who are interested in this functionality."
Not just them, that feature was one of Adobe's biggest selling points for Premiere 64. Then we get folk like the OP who thought it would work wonders, but truth is the only time it works is when your doing a talking heads studio interview, which honestly isn't that hard to edit anyway.
I can see it being very helpful for narration and for big Post houses but that's it. When you do a documentary or dramatic feature it's useless.
It works great for bad actors who are really flat and talk with no emotion though. ;)
>but truth is the only time it works is when your doing a talking heads studio interview, which honestly isn't that hard to edit anyway.
It's not just about editing, although that kind of shot listing is great. Where I'm seeing most of the interest is in archiving and captioning. My own interest is in exporting the text along with the video to interactive web applications.
As you say, it's not perfect and it's still something that perhaps should be on the back burner, but I'm excited by the possibilities.
Do any of you have suggestions for manual transcription workflow? (I don't work in a supported language for the auto transcription). I would love to make notes while I'm reviewing a video (with markers), then be able to print out a list of those markers/notes as a rough index. Clip notes kind of produces something like that when you go through and type comments that are automatically associated with the timestamp. But for whatever reason, when I export to Clip Note/PDF, the "tagged" feature is turned off and I'm not able to print a list of the comments. Also, keyboard shortcuts (space bar to play/pause) don't work when watching the video in the PDF, which makes all the pausing and starting between typing really a pain.