Mac App Transcribes Speech to Text in a Snap
Text-to-speech technology is getting progressively better at recognizing the spoken word, and MacSpeech's latest product, Scribe, throws in transcription as well. Record your words on a portable device, including an iPhone or iPod touch, then hand the audio file over to Scribe. It'll turn it into text in less time than it took you to say it.
Talking to your Mac and watching your words appear on its screen can be an empowering experience, but what about those times you're away from your "iron?" If you have a gadget that records sound files, you can now turn those bytes into text with a new program from MacSpeech.
The maker of the top selling speech recognition software for the Mac, MacSpeech Dictate, has introduced Scribe (US$149), which is designed to transcribe voice files recorded on an iPhone, second-generation iPod touch or digital sound recorder into text.
Like Dictate, MacSpeech claims Scribe is 99 percent accurate. What's more, it's a "smart" program, so the more you use it, the more accurate it gets. In addition, the software, which only runs on Intel Macs with Snow Leopard (OS X 10.6), has a vocabulary editor to train the software to recognize words it's having difficulty recognizing, as well as add new words to its vocabulary.
Turning a voice recording into text is very simple. After training the program with a voice sample from your recording device, you dictate your document into the recorder.
The document will be more manageable if you dictate punctuation as well as text. For example, "See Dick sit. See Jane run." becomes open quote See cap Dick sit period See cap Jane run period close quote.
When you're finished dictating into the recorder, you import the file into Scribe, click Transcribe and watch the program turn the file into text before your eyes. "It converts files into text faster than real time," MacSpeech CEO and Founder Andrew Taylor told MacNewsWorld. "It doesn't play them as if it were listening to them. It goes through them byte by byte very quickly."
Once the voice file is converted to text, the text can be saved as an RTF file suitable for editing in most word processors.
MP3 Crap Shoot
Scribe supports 13 English dialects. In addition, up to six voice profiles can be created for a single user license. Voice profiles tailor recognition to a particular user, making it more accurate.
Six common Mac audio file types are recommended for use with the software -- WAV, AIF, AIFF, M4V, MP4 and M4A. MP3 files may work, but they can be a crap shoot.
"You have to be careful with MP3 because that's a file type with variable compression," Taylor explained. "Some MP3 compressors are good at compacting speech; unfortunately they take out a lot of the markers that speech recognition engines use for distinguishing the different words that are in a file."
The device used to create the voice recordings will influence the quality of Scribe's transcription, he added. In addition to the iPhone and iPod touch, MacSpeech recommends Olympus DS and DM model voice recorders. "Olympus does a great job because one of the reasons that Olympus recorders exist the way they do today is to be a source for transcription products," he said.
"On the Apple devices, it's the same kind of thing," he continued. "They've got very good compressors that don't destroy the speech as far as speech recognition is concerned."
Growth on the Horizon
In addition to the standalone Scribe program, MacSpeech released two Scribe plug-ins for its legal and medical speech recognition products.
Scribe Legal ($99) includes an additional 30,000 legal words and phrases in its dictionary. Scribe Medical ($99) supports vocabularies for more than 54 medical and dental specialties.
In February, MacSpeech was purchased for an undisclosed sum by Nuance, a maker of speech recognition software for Windows-based computers. It's a move that Taylor believes will be very beneficial for MacSpeech, which uses the Nuance speech engine from Dragon NaturallySpeaking in its products.
MacSpeech has been constrained in its growth because it has been tightly tied to revenues. "We've been very careful with our finances, and we're unwilling to commit to initiatives that we can't pay for," he explained. "Nuance has a larger treasury, and if they choose to, they can grow the Mac market much faster with more marketing and advertising."