I wonder what the bandwidth limit would be if you took 44.1kHz audio transmission to the extreme. Something like 56k6 modems?
A .wav file is 16 bits/sample x 44,100 samples/sec ≈ 700 kbps. Thirty seconds of that is about 2.5 Mb, which is a reasonable size for a photo. However, you wouldn't be able to send nearly that much data, since you'll be limited by various kinds of noise.
You're now nyquest-limited to sending fewer symbols per second (baud), but you might be able to use a larger "vocabulary" of symbols.
This was one of the big changes in modem design. The first 300 baud modems used one 1 bit/symbol, but the V34 modems used a bigger symbol set that could send 6-10 bits/symbol (and also sent the symbols faster).