Choose the speaker's impedance, enter the length of the run from amplifier to speaker, and pick copper or copper-clad aluminum. The impedance, in ohms, is printed on the speaker or given in its manual. The length is one way: the cable's two conductors carry the signal out and back, so the current passes through twice the run, and the calculator counts both. The result is the smallest gauge that keeps the wire's resistance within the limit, and how much level that gauge loses over your run.
The limit is the wire's total resistance, out and back, at 5% of the speaker's impedance. Roger Russell's speaker wire table is built on that rule, and Klipsch's support page publishes the same lengths for 4, 6 and 8 ohm speakers. The wire and the speaker share the amplifier's output, so the wire's share is level lost before it reaches the speaker, measured in dB. 50 feet of 16 AWG copper is 0.40 ohm, which is 5.02% of 8 ohms, a loss of 0.43 dB. That just misses: the copper table gives 16 AWG 49 ft at 8 ohms, and for 50 ft the calculator gives 14 AWG. Exactly 5% is a loss of 0.42 dB, and 10% is 0.83 dB.
Lower-impedance speakers need heavier wire. The same 0.40 ohm is 10.04% of 4 ohms, a loss of 0.83 dB, so at 4 ohms 16 AWG copper reaches only 24 ft within 5%, and a 50 ft run takes 12 AWG. The rating is also a nominal figure. A speaker's impedance changes with frequency, and where it dips the wire takes a bigger share. Russell measured a 4 ohm two-way speaker with 1 ohm added in series: it lost about half a dB near 125 Hz and 2.5 kHz, where the impedance was high, and about 2 dB near 300 Hz and 10 kHz, at its minima. Too much resistance changes the frequency response, not just the volume, which is why Russell advises sizing for the lower impedance when the rating is in doubt.
How much of this is audible is less settled than the arithmetic. Jim Lesurf notes that series resistance of a tenth of the load can change level or response by up to about 1 dB, and doubts many people can detect changes under about 0.2 dB. Russell calls the 5% rule conservative and says that with some speakers a loss under 10% would not be audible. The 10% option covers that case. It doubles every reach, so 16 AWG copper goes 99 ft to an 8 ohm speaker.
The size of the copper matters more than its grade. Russell finds that oxygen-free copper (C10200, at least 100% IACS conductivity) has no conductivity advantage over ordinary C11000 copper (101% IACS or better), and that wire type makes no audible difference while resistance stays under 5% of the impedance. Copper-clad aluminum is a different case. ASTM B566 figures for 10% CCA give it about 1.59 times the resistance of copper of the same gauge. Going up two gauges increases the area by about the same factor, so CCA reaches about as far as copper two sizes thinner: in the tables, 14 AWG CCA and 16 AWG copper both reach 49 ft at 8 ohms.
The figures are for solid copper at 20 °C, the same basis as Russell's table. Stranded wire of the same gauge has slightly more resistance, about 2.5% more for 12 AWG, so a stranded run right at a table's limit can need the next size up. In-wall loudspeaker cable falls under NEC 640.21(B) (2023), which points to Article 722: listed Class 2 or Class 3 cable such as CL2 or CL3 (CL2R/CL3R riser, CL2P/CL3P plenum) or a permitted substitute.