The usual answer: you do not need a lot of entropy. 128 bits is enough; if you have 128 truly random bits, you can use them in a cryptographically secure pseudo random number generator (PRNG), which will generate as many random bits as you need, with high speed limited only by local processing power (on a smartphone, PRNG bandwidth will be in megabytes per second, not kilobytes per second).
The continuous collection of entropy is more a fetish than a scientific, rational need. Some say that getting "true" randomness protects you from any cryptanalytic violation on PRNG; but this argument is only saved if you can get fresh evenly random bits (which does not happen in practice: you need to use a hash function to smooth out the collected "noise"), and if you use random bits directly, and not as keys into the encryption algorithm. A stronger case for the continuous collection of entropy can be made in relation to seed storage: the fear that the attacker, having captured the PRNG, can look at its insides, restore its internal state and retroactively guess about random bits that were previously emitted. A good PRNG protects against this. At least you can overwrite 128 fresh bits every second, which is low speed.
However, if you need entropy, camera-to-camera is probably the best source to use, because the CCD detector is very sensitive to thermal noise, and it outputs data with very high bandwidth. A basic one-shot phone camera will easily contain data in megabytes, and even if the phone is inside a hinged-back refrigerator, you will still have a lot of noise (just a thousand bits of noise means over a million pixels, 99.9% are “perfect” , a somewhat ridiculous concept in a $ 400 phone - NASA engineers cannot do this in space research, which costs a million times more).
So just take a picture, hash it with any convenient hash function (like SHA-256) and voila! you have 256 bits of entropy that you use in PRNG. If you're really nervous about PRNG after a while, just take another shot.