@adamhotep @raven667 @Nazani @futurebird Yeah that's a mix of really exhaustive search for optimal compression dictionaries with perceptual models for where to allocate bits that really aren't anything fundamentally fancier than mp3 psychoacoustic model.