I’ve spent hours and lots of money transcribing video from interviews. So when I heard that Adobe Premiere had an automatic transcription technology called “Speech Search,” I was very excited to get the latest version of Adobe Premiere CS4.
Granted, I don’t believe Adobe refers to this technology as automatic transcription technology. I’m the one who called it that. They more see it as a way to search the content of your video, by just looking up specific words. The advantage is those words are synced in time with your video, allowing you to jump to that specific spot in the video.
Well honestly, I think this is a good first try, but I don’t think any of the transcription service companies have anything to worry about. Below is a video of my appearance a couple weeks ago on KQED’s “This Week in Northern California.” I thought transcribing the audio of this video would be a good test. It’s purely a spoken word piece. There are multiple speakers, there’s no distracting noise, and the audio quality is perfect.
Adobe’s “Speech Search” technology is supposed to detect multiple speakers and have good speech recognition. Unfortunately, I think it did poorly on both. It was partially able to identify the speakers and the transcription has absolutely no punctuation. But, to its credit it did get proper names and acronyms correctly. I think it’s a good first step, but they’ve got a ways to go. It will definitely make transcribing easier as you can use this as a first pass before you begin transcribing it yourself.
Once again, Adobe doesn’t sell this as a transcription tool, but rather a search tool. But I believe if it can’t fully transcribe, how good of a search tool can it be?
Take a look at the video below alongside this attached transcription (PDF) created by the Adobe Encoder. I have edited nothing. This is exactly how Adobe transcribed the video. The video is 6 minutes long and it took Adobe Encoder 9 minutes to transcribe the video.