High-speed, locational, phone-to-phone (HLPP) games and apps constitute a provocative class of mobile apps that are currently unsupported on commodity mobile devices. This work looks at a key problem for enabling HLPP: a specific variant of the localization problem in which two phones estimate each other's relative positions in 3D space without infrastructure support. Moreover, position estimates should reflect changes due to the phones' possible mobility.We present a solution for achieving high speed 3D continuous localization for phone-to-phone scenarios. Our basic approach uses acoustic cues based on time-of-arrival and power level. It assumes at least two microphones and one speaker per phone, which is common on new smartphones. Accelerometers and digital compasses assist in resolving ambiguous acoustic-only localization. Continuous localization is achieved with the aid of a loose time synchronization protocol and an extended Kalman filter. Experimental results across a range of motion paths show localization resolution to within 13.9 centimeters for 90% of estimates, and to within 4.9 centimeters for 50% of estimates when the phones are several meters apart.