Voice Recording into text

[techie08]

Does anyone know of any software to transfer Voice recordings into a text document? Our calls are recorded and if they have to be used in a disiplinary, they have to be transfered into a text format which someone has to do manually!

 

Many Thanks

[Steve21]

We have Dragon NaturallySpeaking used within our admin department for them speaking to text, not sure if it can do audio files but guessing as it's just another sound input it would do it.

 

Steve

[chazzy2501]

what about using windows built in transcription? loop the audio back and open word, might work.

[Geoff]

Given the subject I would be worried about the error rate of any speech recognition software. Granted it'll do the bulk of the work, but you should still spend a decent amount of time proof reading what it has produced.

[unixman_again]

Dragon is speaker dependent and I don't think it would work for transcribing. There are apps available but I don't know how good they are or how much they cost. I suspect a typist would be the most accurate transcriber.

[Marci]

An Audio Transcriptionist is the only reliable method of doing this... (my partner does it for a living). You can get an agency involved who will arrange fees for one off files etc. Usually charged per minute of audio to be transcribed.

[dhicks]

Does anyone know of any software to transfer Voice recordings into a text document?

 

As pointed out above, manual transcription might be the only method legally considered reliable enough at the moment. However, Dragon's voice recognition is getting more advanced - it's the same technology behind Apple Siri service and the Dragon Dictation app. The app on the iPad seems to simply do the voice recording, the processing seems to be done on a cloud-based server of some kind. If I understand correctly, that cloud service might well be available as an API for people to build applications on top of, so there might be a transcribe-my-calls website or similar out there already, or you could make one.

[Marci]

Not forgetting that software won't be able to identify one speaker from another, won't handle acronyms unless trained to (which is voice specific), won't detect syballance (deps on audio quality), and will struggle with colloquialisms also. You'd have to get someone to proof and redo, and the total accumulated time would just end up being double the billable time for a pro to do it. Your LA will have a contact that they use for such purposes... or you could contract a freelancer like my better half at £1 per minute of audio.

[plexer]

Sounds like they already use someone to manually do it and from the responses which I agree with looks like it'll have to stay that way.

 

Ben

[Marci]

When our lass started out doing this we tried all the software to see if she could exploit the system to earn easy cash... none of them produced anything usable when dealing with more than one speaker, and on a phone-qual recording of a single voice, had an error rate of over 70%...

 

If going manually from existing staff, find whoever in admin office has highest WPM typing speed, and get them a set of Olympus USB transcription pedals *nod nod*.