- April 14, 2010
- Posted by:
Powerful new search ability uses phonetics to speed media searches
By: Steve Hullfish
– April 14, 2010 –
Avid and Adobe both have technology that allows users to search for specific words in the audio of their footage, but to date, that ability has not been part of the Final Cut Pro arsenal. A new piece of software from Nexidia and AV3 called ‘get’ allows users to quickly search all their media based on actual spoken words within their content, rather than just Finder level data or FCP metadata. Once results are found, they can be used by FCP in the edit. ‘get’ actually searches ANY content that exists on any locally attached media drive, regardless of whether the files have ever been used in any FCP project.
‘get’ also understands FCP’s bin and folder structure and can search information within any column of metadata in an FCP project bin. Search results are displayed in the get Media Browser window. You get a list of media files that contain matches to the input search criteria, grouped logically so the user knows whether or not the files are already being used in an FCP project. Clips can be reviewed with ‘get’s’ built-in player, along with search “hits” marked in a corresponding timeline. The results of the search are accurately located to within tenths of a second. The user can jump to each “hit” for review and can either send the entire clip or file to FCP, or just choose the “hit” that is most relevant (with handles) and send only that portion to FCP. Users have control over a filter which hones the software’s confidence level that it has found a relevant match. So you can ask for high confidence matches to a search and then widen the search to less confident matches.
get indexes media files at over 100 time faster than real time, rendering a one hour file searchable in less than a minute. In my conversation with Nexidia’s Drew Lantham, he claimed indexing speeds of 200x realtime per processor.
For those that have tried Adobe’s Speech Search and have been disappointed with getting consistent results, get is doing something completely different. ‘get’ is indexing based on categorizing speech sounds rather than making specific decisions about words. The technology is actually much closer to Avid’s acclaimed ScriptSync, which simply matches a script with the actual phonetic sounds of the clip, allowing it to be much more accurate than trying to assign a word to a sound without any context. Actually, Avid’s ScriptSync is based on Nexidia technology. Basically ScriptSync just breaks a script into a long series of searches and automatically matches the text in the script to the results of that series of searches.
Also, instead of searching things like proper names, which are very hard to search for with some speech recognition technologies, ‘get’ uses phonetic searching, so exact spelling is not required.
Obviously there will still be a need to do transcripts for many projects. But if the only reason for the transcription is to search for elements of spoken word in the media, ‘get’ can deliver on that need without the transcription.
‘get’ will be available for download in July. Meanwhile, visit http://www.av3software.com to learn more, and pre-register to take advantage of the 10% launch discount. Expected list price is $499.