PRODUCT: Avid’s Phrasefind for Media Composer 5.5
PRICE: $495 (add on)
- For Avid Media Composer 5.5.1 (Mac or PC), Symphony 5.5.1 (Mac or PC) or Newscutter 9.5.1 (PC). Earlier versions are not supported.
PRODUCT: AV3 Software’s GET for Final Cut Pro
PRICE: $249.00 (standalone)
- For Apple Final Cut Pro 6 and 7, with expected support for FCP X.
I thought it appropriate to combine both products in one review as both are based on the powerful Nexidia phonetic search engine and are tools that will essentially change the way we edit. Nexidia, by the way, is well-known in the IT industry for its work in developing phonetic search technologies.
Being on two different editing platforms, the software offerings are not exactly competitors as much as siblings... in rivalry.
To me, dialogue speech recognition is among the most exciting new product categories in recent years. These programs will save larger production companies thousands in transcribing costs (at least in transcribing with timecode), and editors will now be able to search in seconds and “frankenbite” with ease. Independent documentarians and filmmakers will quickly wonder how they ever got along without this technology.
Briefly, the concept for both is simple yet almost magical: type a word or phrase into a search window and voila! You are taken to all incidents of that word (if spoken) in your clips and bins — more or less.
Obviously, mitigating factors such as sound quality, background noise, intelligibility and accents can throw off both programs’ accuracy, but on one (GET) you can dial-in the percentage of accuracy you are willing to accept. But dial it high and you will get less hits, dial it low and there will be a lot of imposters.
While there are many more similarities in how GET and Phrasefind operate, there are as many differences in implementation. The first is, of course, price: Avid’s Phrasefind, downloadable from its Website, is $495, while GET (also downloadable) is currently $249. Both offer a free-trial.
Phrasefind for Media Composer 5.5 is integrated into the editor, while GET is a standalone program that integrates with Final Cut, but can also work outside Final Cut — a boon for story producers, writers, reporters and assistants.
Those familiar with Avid’s ScriptSync will notice one huge operational difference between Avid’s two speech recognition programs — there is no need to enter a script with Phrasefind.
HOW THEY WORK
Both GET and Phrasefind will pre-scan the audio waveforms of clips in bins and build databases.
Additionally, both programs allow new language packs to be added, Currently GET offers Dutch, North American, UK and Australian English, French, German and Latin American Spanish at $169 each, while Phrasefind offers these in addition to Russian at $149 a pop, with more in the offing.
Words typed do not necessarily have to be spelled correctly, just phonetically accurate — as long as it sounds like the word it will search for that word.
While both Final Cut and GET operate independently, they are interdependent. Final Cut should be open so that GET’s clip selections can be exported straight into the editing program.
Once installed, GET will index QuickTime-wrapped media files. You point to what project or file you want indexed. You have a variety of choices — files, projects or entire drives, and make these your “watched” folders and projects. Any new material will be updated automatically and dynamically. You can add as many folders as you want.
It is an extremely fast process... roughly a minute per hour of material. A small clock icon indicates the progress of indexing. (This is true of both programs).
Once indexed, the fun begins! Type in a word or phrase and almost immediately every instance of it will appear in GET’s viewer. A “hit” marker will display how many instances the item appears, and the clip containing these items will have markers attached at each instance. It’s cued to the first instance of the word in question. If you’re searching an actual FCP clip, those markers can be exported along with the clip. Additionally, you’ll find confidence ratings next to each instance ranked from highest to lowest.
Once you have a clip marked, click on the “export” button on the lower right, rename it what you want and it will be added to your FCP projects, with markers conveniently added.
One big difference between the GET and Phrasefind is immediately noticeable: as a standalone program, GET has its own media player — you don’t need to have FCS (Final Cut Studio) or the upcoming FCP X to view and find clips. As such, it makes sense that this could be a great tool outside of FCP. Workstations equipped with GET would allow assistant editor prescreening and paper cutting before edits with the ability to give the editor marked material.
NO FRILLS OPERATION
While not yet as feature-laden as GET, Phrasefind is more fully integrated as an Avid product than the standalone GET is to Final Cut. You’ll find Phrasefind is added as part of a much-enhanced “find” tool for Media Composer 5.5. Avid’s revamped the find feature in a major way to include searching via ScriptSync, timeline text, clip and sequence labels and, finally, Phrasefind.
Once Media Composer is launched, audio waveform indexing begins as with GET. Also, any new material is scanned dynamically as it’s added. It will work on AMA clips, regular clips as well as grouped clips and subclips. Shared storage is not a problem. Phrasefind will reach out to these also, as well as third-party storage. A little raised ball that turns green shows the progress of indexing. Be aware, with AMA linking or importing you must first save the bin before Phrasefind begins indexing.
Once indexed, the Phrasefind bin will populate with clips ranked by accuracy. As of now, unlike GET’s automatic markers for each hit, there is no marking of clips. Instead, clips are loaded in their rank order, so if one clip contains 20 examples of a word, it will appear in the list 20 times, each cued up to the moment before the instance. Not as cool as GET’s marker indexing. However, I am told by Avid that automatic clip marking will be added in the future — they wouldn’t get specific with a time, but said they are constantly updating their tools.
Unfortunately, you can only work out of the search bin; you cannot drag or drop that clip into a new bin. By double clicking you return to the source bin of that master clip, or, if you have “load into monitor” open, that’s where a double click sends it. Phrasefind also indexes in the background and sometimes takes its time. It would be nice to have the ability to coax it a bit and know just where you stand with indexing. The green ball didn’t always tell me... indicating something was indexed when it wasn’t.
Another difference: GET allows you to sort the sensitivity levels of a search: you can dial up or down. Phrasefind does not allow that capability. It’s that simple. No bells and whistles…yet.
Of the myriad challenges today’s editors face, with mounting deadlines, hours of media, less support and bottom-dollar thinking, navigating and finding specific spoken material is one of the biggest time-wasting and headache-making trials we face. Having this technology at our disposal is a joy, and will soon become as essential and commonplace as a matchframe button.
In its V.1.0 iteration, Phrasefind can use a little more tweaking, and I’m sure we’ll see evolution in future releases, but it works.
Is the technology there yet? Yes. Certainly there are the occasional misses, and the accuracy of both seemed to be the same using the same material.
For example, both systems did well with clear voiceover, but messed up when music, background noise or effects were mixed at normal levels. But they both hit much more than they miss and will become an indispensable time- and cost-cutting tool in the editing arsenal.
They will only get better and more feature-enhanced with age. Look for other editing platforms to adopt these tools in the near future.