How can I convert already recorded speech to text?

Question

I am looking for a software that would allow me to input an audio file (.wav .mp3...) and output the text of the audio file.

Does such software exist? I have only found software that converts (live) voice to text, but not audio files to text.

Answer 1

I don't know of any that you can download, but I can make a clunky suggestion.

Google has built such a software, I know that they use it in YouTube and Google Voice.

It's not very good, but you can call your Google Voice account, let the answering service pick up, and play the audio while your voicemail is recorded.

You'll get a transcription of the voicemail in your email.

I only suggest this because you don't have to train Google's engine.

Answer 2

This doesn't answer your question specifically, but what I'd do in your situation is to distribute it as a Amazon Mechanical Turk job. You'll probably get the best results and pay the fewest dollars this way.

Answer 3

eMicrophones makes a commercial product called Windows Speech Recognition Toolkit that adds a lot of goodies to Windows Speech Recognition, including the ability to transcribe *.WAV files (provided they are from you, ie a recording as if you were speaking, not arbitrary voice). From the feature list:

Transcription reads .WAV files from a digital recorder and transcribes it to text. This is meant to be used by the person who has a personal user profile and dictates as they would in front of the computer using good enunciation and punctuation. It is not meant to transcribe multiple voices from a meeting.

Answer 4

You can add VOICENOTE software (free) to google voice. That converts talk to text. Then simply play the voice recording on your speaker directly into the microphone into VOICENOTE.

I plan to use VOICENOTE to transcribe my written notes into text and save in MS Word. Then when I save a document, Windows can search for my handwritten notes on the computer. Otherwise, the handwritten notes can only be saved as a picture. Only the titles are searchaable and you can't find the document words for your search.

Answer 5

There's a software project called 'fanscribe' that breaks up a recording and breaks it up into pieces, distributing them to interested people who can transcribe a portion of it.

https://github.com/fanscribed/fanscribed/blob/master/README.rst