What conversion changes, and what it cannot
Converting re-encodes the audio into a different container and codec. It can reduce file size, make a file playable on a device that rejected the original, or turn a compressed file into an uncompressed one for editing. What it cannot do is add back detail that was already discarded. Converting a 128 kbps MP3 to FLAC produces a lossless copy of a lossy file: the size goes up, the quality does not.
That direction matters when you are choosing a target. Going from lossless to lossy loses a little and is normal. Going from lossy to lossy loses a little again on top of what was already gone, so avoid it when you have the original. Going from lossy to lossless is only worth doing when a tool downstream refuses to read compressed audio.
Sample rate and bitrate
Sample rate can be set anywhere from 8,000 Hz to 192,000 Hz. In practice there are only a few values worth using: 44,100 Hz is the CD standard and the right default for music, 48,000 Hz is the video and broadcast standard and the better choice if the audio will be married to picture, and 16,000 Hz is plenty for speech destined for transcription. Raising the sample rate above the source gains nothing — the extra samples are interpolated, not recovered.
Bitrate applies to the lossy formats and is where the real size-versus-quality decision sits. 128 kbps is acceptable for speech, 192 kbps is a reasonable floor for music, and 320 kbps is the practical ceiling for MP3 where further increases are hard to hear. Leave it empty to let the encoder pick a sensible default.
Tip: Halving the bitrate roughly halves the file size. Halving the sample rate also halves the usable frequency range, which is a much more audible change — adjust bitrate first when you need a smaller file.
MP3 for maximum compatibility, AAC when you want better quality at the same bitrate and your playback is in the Apple ecosystem, OGG for open-format requirements and low-bitrate efficiency, WAV when a tool needs uncompressed input, FLAC to archive losslessly at about half the size of WAV, and M4R only for iPhone ringtones. The conversion itself runs on the server after you export; the uploaded file is deleted immediately afterwards.