The float audio samples are only usefull to operate with them (mix, amplify, etc.), after that must be normalized and converted to int samples before to be played (to avoid overflow).
The precission of 32 float (24 bits of mantissa) is, more or less, the same than 24 int. A float format must be converted always to int (normalize first) to lossless encode.
I never see a source of 32 bit float lossless, maybe is a decode of a lossy format.
__________________
BeHappy, AviSynth audio transcoder.
|