TODO

[jurihock]

• Test AU/VST at least in Logic
• Variable block length in processBlock #2
• Add performance measurements
• Try different FFT backend (PocketFFT, JUCE FFT, Apple)
• Maybe add a configurable gain for individual pitch shifting stages
• Think about advanced automation, maybe something with MIDI chord recognition
• Think about the QDFT integration, e.g. to extract additional features for automation
• Customize the user interface #6
• CI/CD #5

[jurihock]

• It looks like the Logic is reporting the completely wrong (or maybe the max. possible, but still not the configured in audio settings) block size in prepareToPlay and the correct one (configured) in processBlock. So we have to be ready to dynamically adapt the block size as reported in processBlock by reallocating the Core instance.
• I think it's probably more useful to use a stereo layout for both input and output, even if only the left input channel is being processed and extrapolated to stereo output.
• The JUCE build-in FFT operates only on float data, inline, and also "isn't tuned for speed". RIP
• The Apple FFT is a bit odd. The suggested DFT only supports floats and max. length of 4k. The FFT requires a data format, where real and imag parts are "split up" in two memory banks. Actually it still can be a single "packed" memory bank, but there the stride has to be >1 (so I think), which is no more the "best performance" case. RIP