Narration in eLearning

I often use narration in my eLearning courses because I think it adds an additional element of interest, and also because there’s evidence that “presenting words in audio rather than onscreen text results in significant learning gains.”¹ However, I ran into some issues recently that made me reconsider how I obtain the narration audio in the first place. Below is a recap of what happened, and what I decided to do about it.

I was in the habit of asking (begging) co-workers to record audio for eLearning courses. We would usually go into a conference room, record audio using a microphone and Audacity, and then I would import the audio files into Flash, Captivate, or whatever eLearning development tool I was using. This worked well, but I eventually ran into problems. First, if content changed a few days later, we would have to re-record the audio. There’s only so many times you can ask a favor from a co-worker before it becomes an issue. Second, our voices are more inconsistent than we realize; they change depending on the time of day, if you’re getting over a cold, etc. Our audio recordings would often sound like two different people (when it was only one person), especially if we recorded with several days in between sessions. Third, the process takes longer than I would like. If tiny content changes had to be made, it often took an hour or two to work through the process. Finally, I was concerned about turn-over. If I have a co-worker record hours of voiceover work, and then they leave the company, I would probably have to re-record an entire course’s audio the next time changes were made. I decided to look for other options.

I discovered several text-to-speech programs, and they seem like they will do the trick for me. These programs allow you to enter (type in) text and then they output audio files of a person (the computer) reading the text out loud. These programs generally aren’t very expensive, and they output to .wav and/or .mp3 format.

A few text-to-speech programs (most have free demos on their web site):

• TextAloud
• CoolSpeech
• CoolSpeaking
• Verbose
• TextSpeech Pro
• A1 SpeechTRON 1
• Advanced Text to Speech

The programs above come with a standard set of voices, but you can purchase higher-quality voices and plug them into your text-to-speech program:

• The AT&T Natural Voices™
• Cepstral
• NeoSpeech

Voices are generally available for these languages: U.S. English, U.K. English, Spanish, Canadian French, Parisian French, German, Italian, Japanese, Korean, and Chinese.

I did notice some oddities when it comes to pronunciation. Every once in a while the text-to-speech program would stumble on a word, or pronounce a word different that what I wanted. To remedy this, try spelling words phonetically or just try to think differently about pronunciation in general. For example, I had the year "1939" in one of my courses, and the text-to-speech program read it as "one thousand, nine hundred thirty-nine." That’s not what I wanted. So, I changed "1939" to "19 39" (adding a space between the numbers), and it read it properly after that ("nineteen thirty-nine").

I doubt that these programs will ever be able to output speech that sounds as natural as a real (live) person. However, considering how much time they save and how easy they are to use, it is a pretty good solution. Give them a shot if you’re ever in a similar situation.

¹ From e-Learning and the Science of Instruction, by Ruth Clark and Richard E. Mayer, page 83.

Advertisement