SOUND FLIPPER GUIDES
Sample rate and bit depth without fake upgrades
Learn when resampling and precision changes are necessary, why larger numbers do not restore detail, and how the preview preserves samples.
By Singularity Shift Ltd · Updated · Preview guide
Two numbers, two different jobs
Sample rate is the number of samples per second per channel. Bit depth describes the precision of an integer PCM sample. Stereo adds a second channel; it does not double the sample rate. These properties help determine compatibility and the size of uncompressed audio.
For example, one minute of 48 kHz stereo 24-bit PCM contains 48,000 x 2 x 3 x 60 = 17,280,000 audio bytes, before the file header. At the same rate and channel count, 16-bit PCM uses two bytes per sample instead of three. This arithmetic does not describe the variable output size of a compressed codec.
Preserve the source unless there is a reason to change it
Turning 44.1 kHz into 96 kHz does not invent missing frequencies. Turning a 16-bit recording into 24-bit does not recover finer original measurements. A change can still be useful when a production system demands a particular representation, but it should be made for compatibility rather than advertised as recovered quality.
Sound Flipper's conversion paths retain the admitted source rate where the destination permits it. Opus is the explicit exception: output is 48 kHz, with SoXR precision 28 when resampling is needed and a bypass when it is not. That change appears in the resolved plan on the WAV to Opus converter.
Integer PCM and float are not interchangeable labels
Float32 can represent decoder output without forcing it into 16-bit integer steps, and can hold finite values beyond the usual -1 to +1 playback range. An integer file cannot preserve every float value exactly. Sound Flipper therefore refuses float-to-FLAC, float-to-ALAC and float-to-integer AIFF in its basic conversion path rather than silently quantizing them.
Likewise, the qualified FLAC and ALAC paths accept 16-bit and 24-bit integer PCM. A 32-bit integer source should use a supported WAV or AIFF preservation path. Merely choosing a format called lossless does not prove that its particular encoder supports the source precision.
Where dithering fits
Reducing integer precision is a separate signal-processing decision. Dither adds carefully chosen low-level noise before quantization to change the character of the error. It is not a cleaning effect, and should not be added to an unchanged sample-preserving conversion. Read the displayed edit/output plan whenever a precision change is offered.
The FFmpeg resampler documentation describes SoXR precision and dithering controls. See our preservation rules for the current app boundaries and our low-bit tests for why precision checks matter.