I've been exploring the captioning features YouTube offers as a quick way to make a video accessible to people who are deaf or hard of hearing. Unfortunately the automatic voice recognition is horrible. One video I made came back with weird phrases like "you need to copy all mexican bank code" and "indians who helped won't be html dinner." That is definitely not what I said!
What YouTube does well is time code an uploaded script, so the words appear on the screen at the right time. This is how I captioned the video about embedding a Google form and video in a D2L page in my previous post.
When I give advice on creating Camtasia videos, I encourage people to read from a script or at least an outline. However, I just disregarded my own advice and talked pretty much off the cuff when I made the embedding video. I've also heard that people learn better from a conversational voice and I think most people prefer it to when someone sounds like they are reading. Unfortunately, I realized that I use the word "so" to start about half of my sentences.
If I had read from a script, I bet that I would have used less words to get to the same point. As I was typing up what I said afterwards for the transcript file, I noticed there were a lot of "fluff" words in there that weren't necessary. That could be part of what makes it sound conversational though.
In a few spots, I wished I would have added something so I just put it in a callout. I'm not sure how effective this is, so let me know if you have any feedback.
Now that we have Camtasia available for faculty, it would be great to have a transcriptionist as well! I can type about 90 words per minute, so typing up my transcript wasn't extremely time consuming: it took about 1/2 hour for me to transcribe and put in line breaks for a 4.32 minute video. However, the average typing speed is 36 words per minute, which means it would probably take an average typist over an hour for a video this short.
I have used Camtasia for Windows to caption before, but I think YouTube is even easier. Camtasia does integrate with the Windows 7 speech recognition tool. Earlier this week I tried to train the speech recognition to recognize my voice better, but it wasn't even acknowledging that I was talking, so I need to look further into that. Unfortunately, it says it takes about 4 hours of training for it to be reasonably accurate. Why can't the same voice recognition my phone uses (Droid X) be available on computers? My phone is amazingly accurate, and I know Siri on the iPhone is too. Hmmm.