Finding Sections of Audio

[ecdbert]

Hello. I have a long project with repeated sections of audio. I want to find those sections and possibly remove them or turn them to silence. Similart to Find/Replace or Find in a word processing application. Is this possible and how shoudl I do this?

[steve]

If you zoom out so that you can see the entire waveform, repeated sections are often visible by looking at the shape of the waveform.

[ecdbert]

If you zoom out so that you can see the entire waveform, repeated sections are often visible by looking at the shape of the waveform.

This does not quite work. The project is about 4 hours long so to zoom right out makes it difficult to distingish waveform patterns. Besides, I was looking for an automatic way of dealing with those sections.

[kozikowski]

I have a long project with repeated sections of audio.

Describe the show. If you have a performance with bit for bit exact segment duplications, then it may be possible to get a software package to do what you want. If you only have, for example, a performer singing the same words every three minutes, then the project is almost impossible. In that case, the theatrical content and not the bitstream will need to be analyzed. That's much more difficult.

Are you trying to get rid of BBC One's station breaks? If you got the show off-air, the FM noise will kill you. Two station breaks are not the same because the noise is different. If you got it from one of their on-line streams, the digital compression will kill you. Again, two different breaks will have slightly different compression.

Koz

[steve]

I think what you are after is this: http://wiki.audacityteam.org/wiki/Proposal_Audio_Diff but so far it is only at the proposal stage and hasn't been developed yet.

[ecdbert]

I have a long project with repeated sections of audio.

Describe the show. If you have a performance with bit for bit exact segment duplications, then it may be possible to get a software package to do what you want. If you only have, for example, a performer singing the same words every three minutes, then the project is almost impossible. In that case, the theatrical content and not the bitstream will need to be analyzed. That's much more difficult.

Are you trying to get rid of BBC One's station breaks? If you got the show off-air, the FM noise will kill you. Two station breaks are not the same because the noise is different. If you got it from one of their on-line streams, the digital compression will kill you. Again, two different breaks will have slightly different compression.

Koz

Similar to station breaks but not the BBC's and not off air. The waveforms are always the same. Identical bit for bit. Thats why I thought it would be easy. Where can I find such software packages or scripts?

[steve]

If the sections are "bit exact" copies, then that simplifies the matter enormously.

I don't know of any tools that are already available for this job. It is such an unusual task that I doubt that anyone has written one. (Bit exact copies are extremely rare in real world situations).

Writing a scrip to do this would be pretty simple, but performing bit-wise processing of along project is likely to be very slow. Perhaps you could tell us more about the job.
What exactly is the audio that you want to analyze? Where did it come from? What format is it? How long is it? How long are the repeated sections? How is it that they are exact copies?

[ecdbert]

Its a church recording in mp3 format with 'tags' at various points. These are about 20 sec long. I need to remove teh tags.

Thanks for the help!

[steve]

MP3 format will blow it. As soon as you encode anything as an MP3 it is no longer bit perfect.
Do you have access to the original (pre-MP3 encoding) version?