Wednesday, August 8, 2007

Automatic Sync's Caption Sync

I recently tried out an interesting new service provided by Automatic Sync called CaptionSync. We needed to get a program closed-captioned for a feed to PBS stations. Captioning is usually either labor-intensive (not real-time) or poor quality (real-time.) And for labor-intensive, also assume expensive. Not only is it expensive to pay an operator to spot all of the subtitles, but the gear to encode the closed-captioning isn't cheap either.

I found CaptionSync a good alternative to traditional closed-captioning. CaptionSync is a web-based service that essentially works on voice recognition. Now, you need to provide the service a transcript--it can't create a transcript from just audio. But you send CaptionSync the audio, and then it will spot your subtitles for you. After uploading a transcript and an audio file, you wait a couple of minutes and then you get an email with your captions in whatever format(s) you asked for. My cost was $45 for a half hour, and I think that's pro-rated based on $90/hour.

There are a few tricky things with the service. They recommend that the audio you send them only include the voice to be captioned/subtitled. I was sending them a documentary, so I had to make a pass on the film and delete all the audio clips that had no voice or were in a foreign language. Some time, but not too bad. The quality of the spotting was good but not perfect. I had a problem getting one bit of dialogue recognized, maybe because the speaker had an Arabic accent. You can resubmit the job if you aren't happy with the results. For example, I accidentally left in some Arabic voice in the audio track that I sent them, and it totally screwed up the subtitles. So I made a different audio export with no Arabic and then the captions were back on track. I also was able to do some tweaking with the formatting after checking out my results. Again, not a problem to resubmit a job with no extra charge.

Now, the tricky part came in getting the closed-captioning into my program. There is software that will encode closed-captions into a QuickTime movie. It's called MacCaption, and it costs $1000 for a DV version and $2000 for an uncompressed version. No, I didn't add an extra zero. That's one to two thousand dollars just to encode the closed-captioning data into video. Obviously buying the software would blow my budget. So I figured out a work-around. Here's what I did.

I asked CaptionSync for an STL file with my captions. DVD Studio Pro uses STL files to generate closed captions. So I burned that to a DVD. Then I played back that DVD and captured the video as uncompressed. The uncompressed capture records the closed-captioning which is in the vertical interval of the video signal. Now, I don't actually want to use video from the DVD--as it has been heavily compressed and then gone through an analog generation. So I composited the video of the program on top the captured video clip with the captioning. I cut the show in the DV codec, so the program's video didn't overlap the captioning. However, I think an uncompressed clip would normally cover up the captioning because uncompressed video has more lines of video than the DV codec handles. In that case, the solution is to just add a bit of video cropping to the top of the video and the captioning will be revealed. I was working with a TV that could display the captioning while I was doing all of this--to make sure that I had the captions and they were staying in place. Then I mastered to Betacam. I think closed-captioning is supported with DVCam, but I don't know how to make that work.

My work-around was time-consuming, but it didn't cost me any more $$'s. If fit my criteria of cheap with a good-looking master and good (but not perfect) captions. If I was going to be doing this regularly I'd have to think seriously about MacCaption's software encoding of closed-captioning.

No comments: