These are all the things that seem important in a sermon audio management system so far…

I’ve marked things as basic, medium, nice-to-have, advanced feature.

The best approach is probably to get the .basic things going using shell scripts and wrapping other programs, then to work towards integration, and finally towards GUI stuff

The idea here is that it should be really simple to record live, or from existing tapes. If we can handle tapes we can also handle live. But handling tapes means we can solve problems post-record hopefully

This means that somebody should be able to do it while they are working on something else. The distraction of putting in a tape, pressing record on the computer, play on the tape, stop when it changes sides etc is too much.

• record sound from input channel continuously [basic]
• stop when told to stop [basic]
• show input levels [medium]
• allow basic input of metadata (ID3) [medium]
• warn if input level is wrong and the sound is clipping [nice-to-have]

Related software:

• record (part of xawtv), command line utility that handles silence
• arecord is a nice ALSA command line recorder for linux

If we have recorded the sound from a tape we can assume there will be bits of silence that need to be clipped out. The best way to check we have got the right silence is probably to replay the silence after processing. That way the user will not have to do anything or concentrate if it is all silent. But if there is sound that should have been included they will now. This way we can have a fairly aggressive sound filter.

• detect silent sections and clip them [basic]
• save silence separately [basic]
• have a noise reduction filter (band filtering?) [medium]

See Audio Compression for more info on compression formats

• have a set of formats to export to e.g. mp3, spx, vorbis, with standard compression rates [basic]
• for speex we need to resample at 32k before encoding [medium]

• have the ability to upload to a website automatically (selected formats) [basic]
• create a nice simple front end for searching [medium]

Check out Supercast for an example of a streaming audio system that can convert between formats as well

This could be a combination of automatic speech-to-text and manual transcription

• get sphynx or some other speech recognition system working [nice-to-have]
• do text searching on the results [nice-to-have]
• do voice training on existing recordings, remember per speaker [advanced]

(not sure if this belongs here, but just to get the idea down)

Once we have a fairly basic system working, it would be cool to train a few people on how to do tape-to-digital-format conversions. Perhaps we could start by doing the most frequently requested sermons.

It would be good to try and get the optimal quality possible from the tapes - this would mean some processing (e.g. gain, compression, noise-reduction) and possibly a bit of ‘manual’ wave-form editing.

This procedure would need some kind of management (perhaps creating some kind of database at the same time). Not sure if there would need to be some kind of feedback from users/listeners about the quality of the recordings (e.g. if they discover a glitch/plosive etc.) - perhaps recordings can go through a ‘beta’ phase, where intermediate phases can be kept on disk to enable easy editing, then once a recording is ‘finalised/accepted’ (this would include cases where there are problems with the recording, but it is impossible to do anything further to improve the quality via post-processing) these intermediate phases can be deleted. A concern with this approach, is that it might get a bit laborious.