Software Review: Intelligent Assistance Transcriptize

Posted on by Larry

[ This article was first published in the February, 2010, issue of
Larry’s Final Cut Pro Newsletter. Click here to subscribe. ]


I am almost never present during the birth of a new piece of software, but in this case, it happened at our company Christmas Party, last December.

Philip Hodgetts, Greg Clarke, both from Intelligent Assistant, and I were standing around talking technology, codecs, and software — you know, the usual subjects at a Christmas Party — at least for me.

Anyway, I mentioned that when I was writing my book on Adobe Production Premium, I was frustrated by the limitations of their Speech-to-Text Transcriptions. Specifically, there was no way to get them out of Adobe Soundbooth or Premiere except as XML documents, which makes these transcripts virtually useless outside of the Adobe environment.

Greg got a reflective look in his eye and said, “That shouldn’t be too hard to fix.” And Transcriptize was born.

There have been a lot of arguments on the net that the transcripts Adobe Soundbooth or Premiere create are not consistently accurate. Specifically, their transcriptions have problems with proper names and nouns. But, what if you just need a transcript RIGHT NOW and are willing to trade off some accuracy for blinding speed.

In which case, you are half-way home. Adobe can create the transcripts and export them as XML — except that you can’t read them.

That’s where Transcriptize comes in. Here’s how.

For this example, I have an interview with Vint Cerf from 2003 talking about how they are working to redesign the internet (specifically the TCP/IP protocol) so that it can be used in the vast distances of outer space.

In Adobe Soundbooth, I created a transcript and exported it as an XML file. (If you need help in learning how to do this, either refer to my book, or watch the tutorials on Adobe TV.)

Here’s what the beginning of the XML file looks like. The blue highlighted section is the output for the word “what’s” that occurs 1.889 seconds into the video. (XML is like reading a book where there’s only one word on each page.)

After starting Transcriptize, its located in the Applications folder, the opening dialog appears.

I click Choose Transcription XML and select the XML file containing the exported transcript.

I clicked the Identify Speakers button. This smaller dialog says there are two speakers in this transcript. I was one, because I was asking the questions, and Dr. Cerf was the other. This gives me a chance to label the different speakers as part of the transcript, something Soundbooth does not do.

At this point, we have three output choices – ranging from the cool, to the very cool, to the stunning.

* Save Text File.
* Save Spreadsheet
* Save as FCP XML file

Saving as a text file puts your text into a standard text file and labels the speakers. The Adobe software does not punctuate, nor does it put in paragraphs, so you would need to do that as you see fit. The nice thing is that once your XML file is created, it takes only a fraction of a second to convert it to a text file.

For the next option, I’m going to slide the Word Grouping slider down to 5. This creates an artificial paragraph of five words. (You can drag this to any number you want.)

Saving as a spreadsheet creates an Excel spreadsheet that puts each 5 word phrase on its own line, along with the name of the speaker and the time the phrase starts and ends. This makes it easy to do a paper edit of your interview. And, because this is Excel, the person working with the transcript doesn’t need to have Final Cut installed on their system.

But the really stunning part of this software is this. Import the clip you made the transcript of into Final Cut Pro. Select the clip in the Browser and choose File > Export > XML.

Use all the defaults in the Export XML dialog. This exports the XML version of that clip. Keep Final Cut running and the project containing this clip open.

In Transcriptize, click the Choose clip XML button and select the XML file of the clip that you just exported. The clip icon appears in the Transcriptize window. At the bottom, click the Add Clip Markers button.

In the XML dialog, select the project you want this updated clip to be stored – in our case, we’ll use the Transcriptize project.

In almost no time, a new clip is loaded into the Browser with the same name as the original clip. But this new clip has a ton of markers associated with it.

When you load that clip into the Viewer, all those clip markers contain the name of the speaker and the text of what he is saying at that moment in time! (The number of words of text are determined by the setting of the Words slider in Transcriptize.)

Even better, if you edit that clip into the Timeline, then select Edit > Find, you can find for any word contained in the text of a marker!

This means that we can use the power of Soundbooth or Premiere to generate rough transcripts in seconds, then integrate these transcripts into either text files, spreadsheets, or our FCP workflow almost instantaneously.

As someone who does a lot of work with documentaries, this is an amazingly powerful tool! I knew there was a reason we had that party this year…

Note: By the way, I don’t make any money on the sales of Transcriptize – but it was fun to watch Philip and Greg think about how to create this while munching chips in my kitchen….


Bookmark the permalink.

Leave a Reply

Your email address will not be published. Required fields are marked *

Larry Recommends:

FCPX Complete

NEW & Updated!

Edit smarter with Larry’s latest training, all available in our store.

Access over 1,900 on-demand video editing courses. Become a member of our Video Training Library today!


Subscribe to Larry's FREE weekly newsletter and save 10%
on your first purchase.