Adobe Video Text to Speech to be Stored in XMP

Posted by Brooks in Misc on Apr 17th, 2008 | 3 comments

I missed this bit of news coming out of NAB, but thanks to John Nack I’m now caught back up (the benefit of a good feed reader and the echo chamber).

Adobe Text to Speech Tech

I’ve seen lots of talk in the past about video search, deep tagging etc., but nothing that’s effectively realized the potential. It definitely will be interesting to see if Adobe can both build the tech and get the players on board to make it happen. I don’t see anything up on labs at the moment, but hopefully more news and bits will be forthcoming.

For those not aware, XMP is what I would term an RDF media metadata format that’s an important component of the Adobe product ecosystem as well as an open standard that can be used by other vendors. XMP can be injected into binary file formats or saved as an xml sidecar. Adobe provides a free (at least the last time I checked) toolkit for incorporating XMP into your own products.

Metadata’s relationship to search and interoperability is huge and its an arena that’s really still in its infancy, but XMP provides a solid map to largely uncharted waters. I hope Adobe spends more time talking about it. I know its not sexy, but sometimes brains trump our more visceral tendencies.



3 Responses to “ “Adobe Video Text to Speech to be Stored in XMP”

  1. Travis Stoliker says:
    April 17, 2008 at 10:40 pm

    the part you say “this will be huge” — we can already do that man. You can search the Closed Captioning of Videos on google video, try Charlie Rose for instance.

  2. Brooks says:
    April 17, 2008 at 10:48 pm

    @Trav

    Sure and we can already do deep tagging on some sites. However, I haven’t seen any standardization or automation and there are relatively few video sites where this happens. I’m looking for a turnkey solution.

  3. JesterXL says:
    April 17, 2008 at 11:05 pm

    Agreed. Most of the transcoding solutions that do this now, whether as complex as face/image/audio/place recognition or as simple as reading camera metadata are insanely expensive and you have to integrate yourself (meaning, low level coding from scratch). Be nice to get something that is easy for the server-side crew to implement, and the middle tier to expose easily.

Leave a Reply