Adobe Premiere Pro CS4 and Soundbooth CS4 let you convert spoken words into text transcripts, which you edit and search just like other metadata properties. This powerful technology lets you navigate to time locations at which specific words are spoken, helping you better align edits, advertising, and subtitles.  Many Adobe customers are probably very interested in creating videos with metadata that contain transcript information, because those videos become searchable.

Why not offering a service that allows users to upload video and audio material, which we would analyze using the same technology that Premiere Pro and Soundbooth use for their Convert Speech to Text Metadata  feature?
One output format could be a PDF with the original video/audo included at the top followed by the transcript in clear text. If the user plays the video/audo part of the PDF Acrobat could highlight the words in the text as they are spoken ("karaoke").
Another output format could be just the original video/audio file plus the transcript added as metadata.

I played a little bit with convert speech to text metadata using Soundbooth CS4 and found the results often disappointing (the only accurate result was the "Confidence Level", which never exceeded 4% - not even with audio material from The Economist). For that reason I think we need to offer extra UI that allows the user to correct the metadata and the transcript we generate.

Convert speech to text metadata using Premiere Pro CS4:
http://help.adobe.com/en_US/PremierePro/4.0/WS29F692D2-7A49-44db-A109-AE016B766767.html

Convert speech to text metadata using Soundbooth CS4:
http://help.adobe.com/en_US/Soundbooth/2.0/WS29F692D2-7A49-44db-A109-AE016B766767.html