432Hz Batch Converter - Doesn't seem to do proper floating point decoding or something

[NintendoManiac64]

Tested with the AppImage on Linux Mint 20.3 that has been fully updated as of 2023-05-07

Simply put, the 432Hz Batch Converter doesn't seem to be properly decoding floating point or something - it might be truncating everything to 24bit integer and/or incorrectly treating 32bit floating point as 32bit integer instead.

The two main tests I did were to input all 4 of the attached files and have the output set to 32bit WAV and a sample rate of "source"; the advanced settings were configured as:

If you then open the original normalized.wav as well as the "converted" normalized.wav file in Audacity in a single project, invert one of the waveforms, and then mix the two together and then do an Amplify, you'd expect Audacity to report "infinity" but... it doesn't, and instead will let you amplify without clipping by around 90dB or so.

On the other hand, if you open the "converted" wav.wav, aac.wav, and opus.wav files in an Audacity project and do an Amplify, you'll see that it'll say 0dB. This very much implies that floating point data is not being decoded because, if you do the exact same Amplify process with the source wav.wav, aac.m4a, and opus.opus, you will see that Audacity reports that you can "Amplify" by something like negative 48dB.

　

EDIT: This is hilarious - github lets you attach MOV, MP4, and WebM files, but not WAV, M4A, nor Opus? Seriously...

So here's the next-best thing - a ZIP archive containing the aforementioned audio test files hosted on mpv's go-to temporary file host:

• http://0x0.st/HZ6v.zip
• MIRROR: https://web.archive.org/web/http://0x0.st/HZ6v.zip