Out to Launch: Adobe Reveals CS4
In my tests with good mic’d audio, accuracy was in the 90 – 95% range. Certainly accurate enough to help you find certain video sections while editing and a great start for transcriptions (it is free, after all). Editing the text in Premiere Pro is painful; you have to edit a word at a time to preserve the link between the text and underlying audio. Of course, you can copy and paste the entire transcription into a word processor for editing, but then the toothpaste is out of the tube, and you can’t match the text back to the audio.
If you’re a Flash programmer, you can create a Flash presentation that includes the video, audio, and text, and let the viewer search for portions of the video based on the text. For us non-programmers, Adobe promises to make this task easier going forward.
You don’t have to be Warren Buffett to see the value of this feature. All organizations, from corporations to colleges, have countless hours of speech dominated video, whether training, interviews, classes, or the like. The files lie there like a huge, inaccessible lump on a hard drive because there’s no easy way to find what you need.
At a high level, metadata is like losing weight. We all know we could be slimmer, but it’s too hard to do anything about. In one fell swoop, Adobe gave us the ability to add the most important component of metadata without any effort at all beyond pressing the button and waiting for the computer to grind it out.
Anyway, that’s the case I made to Eric and Tim as we drove past Stanford and One Infinite Loop drive in Cupertino. Tim countered that Adobe’s implementation was slower and less accurate than others that he had seen, and that high-quality, near real-time speech to text has been available for years and never implemented in the streaming world.
True, I responded, but Adobe’s is a first-generation feature that will undoubtedly improve over time. And no company other than Adobe controls all the pieces to the puzzle necessary to make speech to text functional and affordable to implement, including Premiere Pro, the Flash Player and Flash Media Server.
Metadata has foundered in the past because data entry was too hard and the benefits very obscure. Beyond the technical cognoscenti, it’s a low profile feature. With CS4, Adobe has made it easy (if not 100% accurate) and brought it to the fore. I could be wrong, but it feels so significant that Apple and Avid can’t ignore it, and their implementations, with full view of the limitations and mistakes of Adobe’s, should be superior, and will push the overall pace of innovation.
That’s my take anyway. Gotta go; have a tee time to make.