The Soundbooth CS4 beta in Adobe Labs has a great audio transcription feature to be integrated into CS4 apps, which was discussed here previously. I got very good results transcribing a lecture and podcast (a good recording in American English with clear diction, and no speakers talking over each other). Here's one way to do this, from posts by Scott Carver:
1. Expose the Metadata panel (Window->Metadata)
2. In the Metadata Panel, find the section called Speech Transcript (after the File and Clip sections). From here, you can transcribe, make corrections to your transcription, play back individual words, etc. In addition, the search box at the top of the metadata panel will search for words in the transcript as well as text in the metadata of the file.
Update: Beet.TV reports that Youtube is doing transcription now too, along with Blinkx and others, so be careful because video search is getting a whole lot easier:
'The search box is found on YouTube You Choose page. (It's on the lower left area of the page.) Visitors can enter words such as "Iraq, "Global Warming," and "Free Trade." Also, putting your cursor over the timeline, you can see the transcription.'
Update 2: In an August '08 EventDV article, Jan Ozer muses that this will be a big feature. Others seem to concur -- in an NAB tidbit, Beet.TV and an Adobe rep talked about how Video metadata is key to web future. And as noted in NAB 2008 on Adobe TV, Hart Schafer did a demo of Production Premium CS4 showing off this feature (better viewed at Adobe), which works a bit differently than Avid ScriptSync (see the Avid overviews).