Free timestamped audio transcription (speech to text)?
speech-recognition
Solution
CMUSphinx can do this, just run Aligner demo from the sources. The latest sphinx4 tutorial is available in sphinx4 tutorial. In the summer 2014 CMUSphinx has integrated long audio alignment functionality in sphinx4 trunk.
It is totally wrong but unlike voicebase it captured at least some words. Probably it's easier to transcribe it manually and later assign times. CMUSphinx can also do this with audio-to-text alignment.
There are quite many issues with decoding accuracy caused by accents and shorttalk, but there are things you are responsible for too. For example you should never convert telephone audio of 8khz bandwith to mp3. Such conversion drastically reduces the accuracy of the automated speech recognition and the overall recording quality. You could use other speech-oriented lossless codecs for compression.
Other question discussing the subject, listing some recommendations.
Problem
Is there any free service/program that will transcribe audio to text and preserve timestamps? I want to transcribe: This is the demo link of VoiceBase (a 16m conversation between 419 scammers; the scammers speak with heavy accents, so I'm not complaining about the quality of the transcription). Voicebase DOES transcribe it, but the transcription is a single block of test, and is not timestamped. I also tried the Google voice "trick" (record the file into voicemail and have Google transcribe it 3 minutes at a time), but, again, the transcription is a single block of text, not timestampped. I am aware of Sphinx, but couldn't get it to work. I realize I could use subtitleeditor to create a manual trascription, but I'm looking for an automated solution. Ideally, I'd like to take a timestamped machine transcription and hand edit it.