Automatic audio transcription is not worth it
Automatic audio transcription is not worth it. And see: if it were worth, why would the Written World (and other reputable companies) rely on transcribers to always pay for the service? In the market system in which we live, no company would let this opportunity slip if it really paid off.
Many attempts have already been made in this direction by IBM, Google and other large companies. We also very much wish that this dream of transcribing audios automatically becomes a reality.
Why is automatic audio transcription not worth it?
The proposal of automatic audio transcription is not worth yet because the software needs to convert the human voice (recorded or live) into quality text. Otherwise, the time needed to revise it after conversion turns out to be higher than what you would spend typing it manually.
And more! The texts produced by these software are very dangerous because the post-conversion review should be much more careful, since they never return words that do not exist. That is, if the pronunciation is not well done, the software will return the text that most closely matches what has been said – as with the text-to-speech converters of smartphones. Imagine the amount of errors that can result from more extensive audio …
It is true that in informal activities these software can help. But if the purpose of transcription of your audio is to generate protocol documents, especially if the content has specific terminologies, do not waste your time. Transcribe it manually and make sure everything is out of control. Otherwise, you run a lot of risk of harming your document.
Experiments with automatic transcription software
As providers of audio transcription (and de-duplication) service, of course we’ve done a lot of experimentation until we come to that conclusion. And that’s why we wrote this article.
We have even had the experience of passing the texts converted by the software to our proofreaders, but, of course, also for financial reasons, this did not work, since the text proofreading service is more expensive than the transcription service.
IBM was the first to do voice-to-text conversion. Assuming that the human voice differs from person to person, when ViaVoice was installed on the desktop, it was necessary for the user to do voice recognition. The user had to read, aloud, all the texts that the software requested, until the voice recognition was made.
There was no progress unfortunately
I can not even see that it’s been 10 years! Today Google and other manufacturers are already converting from voice to text without the need for voice recognition. But unfortunately, the conversion features are exactly the same.
If 10 years after ViaVoice conversion technology remains the same, despite the efforts of the powerful Google and other large companies, the hope that software in the market that actually fulfills the role of converting voice to text with quality still It is sparse.
So if you can not keep waiting until the software is actually developed, we suggest you learn how to manually transcribe or hire the service of those who already do it professionally.
With this post, we hope to have contributed so that you do not have to try out the various features of automatic conversion available, in the end, like us, also realize that automatic audio transcription is not worth it.