Wow. Filesizes

[ssj71]

yes, it makes sense to mix in 24 or 32 bit because you lose dynamic range through processing, but for export and playback, 16 bit is plenty.

[sadko4u]

Yes, that's a great article. Remember that every change of sample rate that is not multiple is lossy.

Even if it is multiple it's lossy. Because for downsampling you have to apply low-pass filter first before doing resampling.

[ssj71]

Even if it is multiple it's lossy. Because for downsampling you have to apply low-pass filter first before doing resampling.

These are tiny details, but theoretically its only lossy if there is content above the nyquist frequency of the new rate and only that higher content gets lost (which is inaudible if downsampling from 96k). In practice though it also depends on the quality of your low pass filter a perfect "brickwall" filter takes an infinite amount of time to apply, so you tradeoff computation speed with quality.
I think I'm preaching to the choir though.

[chaocrator]

why bother about audio files size in the 21st century, when the storage is REALLY cheap?

[sysrqer]

why bother about audio files size in the 21st century, when the storage is REALLY cheap?

Not every one has spare cash or live in a country where it is cheap.

[Lyberta]

why bother about audio files size in the 21st century, when the storage is REALLY cheap?

Bandwidth is not so cheap.

[tramp]

Flac uses a lossless compression, and is what you want if you wish to retain all original audio information.

Know it now. I will remain with flac or wave now i know that.

You need to know also that flac didn't store values above -1/1, so, when you've run into clipping and recorded in flac, you cant restore the sound by simply reduce the volume or normalize.