I did a bunch of math today and estimated that I could maybe squeeze 1.6 kbps out of a cassette tape if I use like... seven different frequencies and a fast fourier transform.
then I checked wikipedia and found out that in 1978 some folks managed to get 1.7 kbps with just two frequencies and a high-pass gate.
https://en.wikipedia.org/wiki/Kansas_City_standard#Enhancements
The MSX by default supports both a 1200 baud variation of the standard with the same bit encoding as Acorn's, and a 2400 baud variant which doubles the audio rate — a "0" bit is one cycle of a 2400 Hz wave and a "1" bit is two cycles of a 4800 Hz wave. Unlike Acorn machines, the MSX uses two "1" stop bits in addition to one "0" start bit, so the effective rate at 1200 baud is approximately 873 bits per second, and the effective rate at 2400 baud is approximately 1,745 bits per second.
for context that means to download a 250 KB file (a pretty common size for a ZZT game) would take 20 minutes of tape, and that's if it loads correctly the first time. I think people used to save everything twice in case of errors, so that's 40 minutes. for 250 KB.
oh wait if you zip the file you can get it down to about 50 KB, so that's closer to 4 minutes for the first read and 8 for two reads. and I think it's safe to assume my target audience has an unzip program. so... this might be doable. hmm.