Is is possible to do automatic upsampling of MP3's? I mean, if for instance a MPEG file is 32kHz, the output can be 44,1kHz anyway.
There's been some interest in whether something like this can be done as part of the decoding process,
There are several Linux audio drivers that do this when faced with hardware that will only operate at one (or a few) sampling frequencies. They simply replicate (or mabye try to linear interpolate) samples such that over a small time window the start and end samples are properly timed. Since it's the final output stage, it seems to be a sufficient solution.
While linear interpolation or replicating samples is okay for small computer speakers or cheap headphones, even non-audiophiles like me can tell the difference between poor resampling and good resampling, when I've good good speakers to reproduce the sound.
There are a lot of free resampling algorithms out there, but none of them are drop-in libraries that are easy to use in an existing program (like libmad is). A Sourceforge project was recently started to address that problem. I think that their API is actually very nice - and it would make it possible for tools like libmad to resample "on the fly". So far the source code they've made available just has a single implementation based on linear interpolation, but more methods are coming.
http://libsample.sourceforge.net/
- Dominic